# Orkindo — Enterprise AI agent platform ## Build the Brain of Your Company Build internal apps, run autonomous workflows on schedules and events, and talk to the same agents on WhatsApp, Slack, or Teams. ## One Brain Instead of Scattered Systems Your company's knowledge is spread across databases, documents, tools, and processes that never talk to each other. Orkindo connects to all of them and becomes the single brain that centralizes, orchestrates, and governs everything. ### Knowledge scattered across disconnected systems → One brain connected to everything Databases, documents, internal tools, and business processes all live in separate places. Orkindo connects to every one of them and brings the whole company into a single layer of intelligence, so the same brain that reads your order database also knows your returns policy. ### Information locked behind systems → All your data, centralized and ready to answer Orkindo digests your databases and documents into one centralized brain. A manager types "What were our top products last week?" on WhatsApp and gets the table straight from the sales database, source-cited and in seconds, no SQL or analyst handoff required. ### Manual processes nobody orchestrates → Orchestrated, autonomous action Reports, alerts, and processing that someone runs by hand never stop. The brain orchestrates these as autonomous workflows: a daily production summary pushed to managers every morning, an alert the moment inventory crosses its threshold, all on schedules or events with no human in the loop. ### No control over who reaches what → Governance on every connection A central brain with access to everything needs governance. Role-based permissions, audit logs, and full traceability on every database, document, and tool, the discipline that lets a bank open its compliance policies to every employee while every single lookup stays logged for regulators. ## One Brain for Your Entire Company Orkindo connects to all of your systems, digests everything into one place, and acts on it. A single enterprise brain built on three pillars: centralized, orchestrated, and governed. ### Connected to Everything Orkindo plugs into your databases, documents, internal tools, communication channels, and business processes out of the box. Every source of company knowledge becomes part of one brain, with nothing left in a silo. ### Centralized Intelligence The brain digests all of that data into a single layer anyone can reach. Ask in plain English on WhatsApp, Google Chat, or web chat and get tables, charts, and source-cited answers from across every connected system, no SQL, no analyst handoff. ### Orchestrated Action The brain does not just answer, it acts. It orchestrates autonomous workflows across your connected systems, on schedules or when events occur, coordinating data and tools to get work done with no human in the loop. ### Governed by Design A brain with access to everything needs control. Role-based access on every system, database, and knowledge source, full audit logs, token usage tracking, and traceability on every action the brain takes. ## Every connection makes it smarter Each flow, agent, database, document, and tool you connect adds context. The Orkindo assistant uses all of it when it helps you build the next flow, app, or agent, so the tenth one takes a fraction of the effort of the first. ### Connect Plug in a database, a document library, a tool, or a channel. Each one becomes part of the brain's shared context, available to every agent and every builder. ### Context Schemas, documents, past runs, and the agents already in place are known to the assistant. Nothing has to be re-explained, and every new piece makes the picture more complete. ### Build Ask for a new flow, app, or agent and the assistant proposes it from what it already knows: the right tables, the right tools, the right channel, the right tone. **In practice:** Connect your CRM database and the support inbox. When you later ask for a churn-alert agent, the assistant already knows your customer fields, your ticket categories, and which channel your team actually reads. ## Everything the Brain Needs to Run Your Company The connections, intelligence, and governance that let one brain centralize your data, orchestrate autonomous work, and stay under control on every channel your teams use. Each card below has a full section further down the page. ### Conversation With Your Data Ask your databases questions in plain English, "How many units shipped yesterday?", and get answers back as tables and charts, with follow-ups that keep the context. Works across PostgreSQL, MySQL, Oracle, SQL Server, BigQuery, Snowflake, and 13 engines in total. Agents connect through a read-only role the platform provisions for them, so they can never write to production. ### Conversation With Your Documents Chat with manuals, policies, and procedures the same way. A field technician asks for the lockout procedure of a specific pump station and gets the sourced step-by-step on the spot. Ingest from PDFs, Google Drive, OneDrive, or S3, keep every collection in sync automatically when the source changes, and every answer cites the original document. ### Complete Internal Apps Build complete internal applications connected to the entire Orkindo ecosystem. Create systems that run closing meetings, dashboards with insights from client meetings ingested through Google Meet, operational portals, and other custom experiences, using the same data, workflows, and governance controls as the platform. ### 222 Ready-to-Run Assets Start from a working blueprint instead of a blank file. 167 templates, 34 ready-to-run workflows, and 21 reusable tools across 17 categories, from sales, support, and operations to finance, HR, legal, security, and analytics. Import one, point it at your data, and it runs in minutes. ### Scheduling and Event Triggers Run workflows on a schedule, on webhook events, or on incoming messages from WhatsApp, Telegram, or Google Chat. The daily production summary lands in managers' chats every morning without anyone asking; the stock alert fires the moment the threshold is crossed. ### Every Channel Your People Use Reach customers and teams on WhatsApp, Telegram, SMS, Messenger, Instagram, LINE, email, Google Chat, a chat bubble on your own website, your own systems through the API channel, or a realtime voice call in the browser. Same agent, same answers, every channel. ### 100+ LLM Providers Power your workflows with the best model for each task. OpenAI, Anthropic, Google, Azure, AWS Bedrock, or local models via Ollama, switch without changing agent logic. ### Code Assistants Under Governance Run Claude Code, Codex CLI, and OpenCode inside the platform instead of on unmanaged laptops. Terminal and chat side by side, session continuity for long tasks, workspace files synced to a storage connection you control, and every session in the audit log. ### Access Control and Governance A nested role hierarchy with 180+ individual permissions, per-record allowlists, MFA enforced by role, and a full audit log. Agents reach databases through read-only credentials and only the workflows, tools, and knowledge collections they were explicitly granted. ### LLM Cost Tracing and Observability See every LLM call your workflows make. Track tokens, cost per workflow and per model, average duration, and drill into individual traces to debug slow or expensive runs. ### Workflow Eval Suites Score each workflow against a fixed set of test cases. Run the suite after every prompt or model change and catch regressions before they reach production. ### MCP Servers and Custom Tools Extend agents with built-in MCP server tools like Playwright, Fetch, and Sequential Thinking, or register your own custom tools. Chain tools together and compose workflows out of workflows. ### Skills, Plugins, and Marketplaces Package instructions as Skills, bundle Skills with MCP servers as Plugins, and install both from Git-backed marketplaces: the official Claude Code and Codex catalogs, or your own private repository of approved packs that admins curate. ### CLI, Git, and MCP Both Ways Pull workflows, tools, and templates to a folder with the orkindo CLI, edit them in your editor, version them in git, push them back. Connect Claude Code or Cursor to your workspace over MCP, and give your own agents MCP servers of their own. ### Your Own Dedicated Instance Every customer runs on an isolated instance: own domain, own PostgreSQL, own Redis, own containers, own network. Run it in our cloud, self-host it, or deploy it in your data center, and reach private databases behind your firewall without opening a port. ## Every Answer Comes From Your Own Systems A generic chatbot knows nothing about your company. Orkindo agents answer from the databases and document libraries you connect, and only those, with a query or a citation behind every answer. ### Conversation with your databases Register a database once and every agent you allow can query it in plain language, on any channel, returning tables and charts instead of SQL. - **13 engines, one interface**: PostgreSQL, MySQL, Oracle, SQL Server, BigQuery, Snowflake, ClickHouse, DuckDB, and more: 13 engines in total, all queried the same way from the same conversation. - **Verified before it is saved**: A connection cannot be saved until the platform has actually reached the database. A wrong credential fails on the settings screen, not in a conversation at midnight. - **Learns your schema and your vocabulary**: Teach the agent what "active customer" or "net revenue" means in your business and which tables hold what. Training data turns a generic SQL generator into one that knows your reporting conventions. - **Read-only by construction**: The platform provisions a SELECT-only database role for agent access and never swaps it for the admin credential. An agent can answer any question and cannot change a single row. - **Metabase included**: A full BI tool ships with the platform. Every database you register appears in Metabase automatically, so the data behind a chat answer is one click from a dashboard. ### Conversation with your documents Knowledge collections turn manuals, policies, contracts, and procedures into a library that answers questions and stays current on its own. - **Any source you already use**: Upload PDFs directly, or connect Google Drive, OneDrive, S3, or MinIO. Collections are scoped by permission, so the HR policy library and the engineering runbooks are never the same collection. - **Updates itself**: Change a file in Drive or drop a new version into a bucket, and a webhook re-indexes the collection. Nobody re-uploads a policy, and nobody answers from last quarter's version. - **Graph retrieval, not just similarity**: Retrieval is graph-based: the platform maps the entities and relationships across your documents, so a question that spans three procedures gets one coherent answer instead of three loosely matching paragraphs. - **Every answer cites its source**: Answers point back to the document and the passage they came from. An auditor, a technician, or a regulator can check the claim in one click. **In practice:** A regional manager asks on WhatsApp, "Which stores missed target last week, and what does the incentive policy say about it?" The agent runs a read-only query against the sales database, pulls the policy from the knowledge collection, and replies with a table, a chart, and the cited paragraph. ## Ship Workflows With Their Own Branded UI Every workflow can come with a custom screen, an intake form, a dashboard, a multi-step wizard, embedded on any internal page. No separate frontend project, no extra hosting, same governance and audit trail as the workflow itself. ### Intake & Approval Forms Give a workflow a branded form to collect customer requests, vendor onboarding details, or internal approvals. Submissions flow straight into the workflow, with validation and routing built in. Includes: - Validated fields - Routing rules - Shareable links ### Internal Dashboards Turn a workflow's output into a live dashboard for ops, sales, or leadership. Charts and tables refresh from the same data conversations your agents already run, no separate BI tool required. Includes: - Charts and tables - Live data refresh - Embed on any page ### Multi-Step Wizards Walk users through a guided process: a configuration setup, a quoting flow, a compliance checklist. Each step can call agents, query data, or trigger downstream workflows. Includes: - Guided steps - Agent-powered logic - End-to-end audit trail ## Autonomous Agent Workflows for Every Team Automate the repetitive work behind every team, from field operations to production to back-office reporting. The same workflows are reachable on demand via WhatsApp, Google Chat, or web chat, and when a project needs more than one agent, promote it to an Agent Organization. ### Field Operations Workflow Agent workflows that surface equipment manuals, maintenance history, and troubleshooting steps automatically, and respond to field technicians on demand via WhatsApp. Instant answers on-site, no office trip required. Connected to: - Equipment manuals - Maintenance logs - Troubleshooting guides ### Production Floor Workflow Agent workflows that monitor inventory, enforce procedures, and watch quality metrics in the background, and answer production staff over WhatsApp or Google Chat without leaving the floor. Connected to: - Inventory systems - SOPs & procedures - Quality data ### Operations & Reporting Workflow Agent workflows that generate reports, query data across every company system, and deliver insights on schedule, also reachable on demand via WhatsApp or Google Chat for ad-hoc questions. Connected to: - All databases - Company knowledge - Reports & KPIs ## Chain Workflows Into Pipelines One workflow is a building block. Orchestration turns the workflows you already built into a pipeline: first run A, then feed its result into B and C at the same time, and when both finish, run D. No glue code. You draw the pipeline on a visual canvas. Drag workflows in as nodes, connect them, and every connection passes one step's output as the next step's input. Steps that do not depend on each other run in parallel, and a merge step waits until every branch feeding it has finished. ### A Visual Pipeline Canvas Drag your existing workflows onto the canvas and wire them together. The graph is validated on save, so a broken chain, a loop, or an orphaned step never reaches production. ### Fan Out, Fan In Independent steps run in parallel, and a merge step starts only when everything feeding it has finished, with all the upstream outputs in hand. Extract, transform, load, without writing the coordination. ### Cron, Webhook, or One Click Run the whole pipeline on a schedule, let an external system start it with a single webhook call, or trigger it manually. Your coding assistant can build and run one over MCP too. ### Watch Every Step Each run shows the live graph with every node color-coded from pending to running to done, one click from the full execution log of any step. Past runs keep a frozen snapshot of the graph they ran with. ## Hire a Team of AI Agents Real work is rarely a single question. It is a project: plan it, split it, draft, review, approve, ship. Agent Organizations let you stand up an org chart of agents, a lead and specialists, that carries a project from idea to done. Think of it as hiring a small department that never sleeps. You set the mission and define who reports to whom; the team picks up work from a shared board, does it, and reports back, pausing for your sign-off wherever you decide it matters. ### An Org Chart of Agents Each agent has a role, a manager, and a lane: a lead, specialists, analysts. Every agent is powered by a real workflow or development assistant underneath, so it does the work instead of just talking about it. ### A Shared Task Board Work moves through clear stages, from backlog to ready, in progress, review, and done, updating live as agents pick up and finish tasks. You always see who is doing what and what is waiting on you. ### Human Approval Gates Gate sensitive work so nothing runs until a person says go, and route finished deliverables to a reviewer who approves or sends them back. The team iterates on that feedback on its own until you are satisfied. ### Budgets and Schedules Cap spend at the team and per-agent level, and give agents a schedule so recurring work happens without anyone kicking it off. Every run logs its reasoning, its actions, and its cost. ## An Ecosystem, Not a Feature List Agents on Orkindo install capabilities the way developers install packages: from a public marketplace, or from your company's private catalog of approved packs. Admins decide which sources a workspace may pull from. ### Skills Reusable instruction packs that teach an agent how to do one thing well: how your company writes a quote, how a claim is triaged, how a report is formatted. Write one, attach it to any agent, update it in one place. Example: "Draft a quote in our house style" ### Plugins Bundles of Skills plus the MCP servers they need. A plugin for invoice processing ships the instructions and the tools together, so installing it is one action, not a setup guide. Example: Skills + a Fetch server + a Playwright server ### Marketplaces Git-backed catalogs you sync and import from. Use the official Claude Code and Codex marketplaces, or point a workspace at your own repository so only reviewed packs are installable. Example: github.com/your-company/approved-packs ### 222 assets, ready to run The template library is not a demo. Import a template, point it at your data and your channel, and it runs. Every asset is also a starting point the Ask AI copilot can adapt to your case. - 167 templates - 34 ready-to-run workflows - 21 reusable tools 17 categories: Development, Operations, Communication, Support, Analytics, Sales, Marketing, Content, Data, Finance, HR, Quality, Research, IT, Legal, Scheduling, Security ## From Idea to Live Workflow in Three Steps Build the agent in the browser or in your own editor, deploy it to any channel, and watch it in production, with the numbers to prove it still works. We deliver your first agent in production in up to two weeks, and usually in one. ### Step: Build Write your workflow in Python inside the Builder Area, with an Ask AI copilot that learns from the workflows, tools, and data already on your platform, or pull the workspace to your machine with the CLI and work in your own editor. Test live in the side chat panel and watch every LLM call in the integrated trace view. - Python editor with Ask AI copilot - CLI and MCP for your own editor and assistant - 222 templates, workflows, and tools to start from ### Step: Deploy Save your workflow and deploy it instantly, schedule it, trigger it on webhook events, or expose it on WhatsApp, Google Chat, Telegram, or any of the 15 channels, including Slack, Microsoft Teams, and Discord, down to a voice call in the browser. No infrastructure setup; it's live in seconds. - Scheduled and webhook-triggered runs - WhatsApp, Telegram, Google Chat, and more - Zero infrastructure setup ### Step: Monitor & Improve Inspect every execution in real-time logs, track cost and token usage per workflow with LLM tracing, and score new versions against fixed test cases with eval suites before promoting them. - Execution logs with full input and output - LLM tracing with per-workflow cost - Eval suites for regression testing ## Real Code, Real Files, Your Own Tools The buyer signs and the developer decides. Orkindo is built for that developer: workflows are Python, the workspace is a folder, and your coding assistant already knows how to talk to it. ### Workflows are Python, not a locked canvas Write agents in the Builder Area with a full editor, an Ask AI copilot that knows your existing workflows and schemas, a live test chat, and an integrated LLM trace panel. When a canvas is the right tool, Orchestration gives you one, on top of the code. ### A CLI that treats your workspace as files Pull workflows, tools, and templates into a local folder, edit them in your editor, version them in git, review them in pull requests, and push them back. Login is device-style OAuth from the terminal, with no credentials stored on disk. ### Orkindo is an MCP server Point Claude Code, Cursor, or any MCP client at your workspace and it can list, read, create, and update workflows, tools, knowledge collections, and contacts from the terminal. Build an agent from your coding assistant without opening a browser. ### And an MCP client Give agents MCP servers of their own: Playwright, Fetch, and Sequential Thinking ship built in, and you can register any server you run. Tools are versioned, chainable, and composable into workflows of workflows. ### Coding agents under IT governance Claude Code, Codex CLI, and OpenCode run inside the platform with session continuity, workspace sync to a storage connection you control, and an audit trail. The answer to Shadow AI is not a ban, it is a governed place to run the same tools. ### Switch models without touching logic 100+ providers behind one interface. Move a workflow from a hosted frontier model to a local one via Ollama, or the other way, by changing a setting, then prove nothing regressed with an eval suite. ### Your workspace as a folder 1. `orkindo login` — Device-style OAuth login, nothing stored on disk 2. `orkindo pull` — Workflows, tools, and templates land in a local folder 3. `git commit` — Edit in your editor, review the diff in a pull request 4. `orkindo push` — Push the reviewed change back to your instance ### From your coding assistant Connect Claude Code or Cursor to your workspace over MCP and ask for a new workflow in plain language. The assistant reads the tools and collections already there, writes the workflow, and saves it to the instance. ## One Brain Connected to Everything Plug the brain into your databases, communication channels, AI providers, and enterprise tools out of the box, so every source of company knowledge feeds one centralized intelligence. ### Databases Agents query any database with natural language - PostgreSQL - MySQL - Oracle - SQL Server - BigQuery - Snowflake - ClickHouse - DuckDB ### Channels One agent on all 15 channels your customers and teams already use, from Slack and Microsoft Teams inside the company to WhatsApp and Instagram outside it - Slack - Microsoft Teams - Google Chat - Discord - WhatsApp - Telegram - SMS - Messenger - Instagram - LINE - Email - Website Widget - API Channel - Realtime Voice - Webhooks ### AI Providers Power agents with 100+ LLM providers or local models - OpenAI - Anthropic - Google - Azure - AWS Bedrock - Ollama - Together AI - Cohere ### Tools & Frameworks Extend agent capabilities with powerful AI frameworks - Metabase - MCP Protocol - LangChain - AutoGen - CrewAI - PydanticAI - LangGraph ## Humans and AI, One Support Desk Turn every channel into a customer support desk. Bind a channel to an inbox and every incoming message becomes a conversation your human agents can read and answer, side by side with your AI workflows. ### Bot First, Humans Take Over Your workflow answers first while agents watch the conversation live. One click takes over the thread, mutes the bot, and the customer never changes channels. The workflow itself can escalate too, the moment a customer asks for a person. ### Human Agents Only Some conversations should never see a bot. Route those channels straight to your team, and every message waits for a person from the first word. ### Shared Inboxes One inbox per channel account, with conversations moving through open, pending, and resolved, visible to every agent in real time. ### Teams and Auto-Assignment Group agents into teams and let round-robin assignment distribute new conversations automatically, so no one is overloaded. ### Labels Color-coded tags like billing, complaint, or VIP that agents attach to conversations and filter by, shared across all inboxes. ### Canned Responses Reply faster with shortcuts: type a slash and a short code, and the full response drops into the reply box. ### Beyond the inbox One support desk that replaces a standalone help-desk product, on the same instance as your agents and your data. - **One contact, every channel**: The same person on WhatsApp, email, and the website widget is one contact record, with identities mapped across channels, tags and groups for segmentation, and a schema-driven merge that keeps a full audit trail. - **Campaigns**: Send a one-off broadcast to a contact group or a tag, decide per campaign whether replies go to the bot, a human, or an inbox, and track deliveries and replies as they come in. - **Awaiting-reply queue**: Conversations waiting on your team surface in one queue, so the oldest unanswered customer is always visible and nothing quietly ages out. - **An agent workspace on any device**: The support workspace installs as a mobile app with push notifications, so a human takeover on WhatsApp can happen from a phone, not just a desk. ## Control You Can Show an Auditor "Role-based access control" is what every vendor says. This is what it means on Orkindo, control by control, so a security review can score it instead of guessing. ### A role hierarchy, not a flat list Roles nest, and a user's permissions come from the role they are actively using in the session, not from the union of everything they were ever granted. 180+ individual permissions cover every workflow, tool, database, collection, channel, and admin action. ### Row-level allowlists Beyond role permissions, individual records carry their own allowlist of users and roles. A workflow, a database, or a collection can be visible to exactly the people who should see it, and nobody else. ### MFA enforced by role If any role assigned to a user requires multi-factor authentication, that user gets it, no exceptions. Enrollment, challenges, and resets land in a dedicated MFA audit log. ### Read-only agent credentials For every database the platform mints a SELECT-only role, and for object storage a list-and-get-only user, and connects agents through those. The admin credential is never handed to an agent. ### An audit log with real coverage Workflow edits, executions, data access, channel messages, human takeovers, approvals, credential changes, and MFA events, each with who, what, and when. It is what a bank shows its regulator, not a marketing checkbox. ### Spend caps at every level Token budgets per team and per agent, cost tracing per workflow and per model, and usage reports per role. An agent that runs away hits a ceiling, not a credit card limit. ### Sandboxed execution Each workflow runs in its own isolated Python environment with its own dependencies. One workflow's package cannot break another, and nothing runs with more access than it was granted. ### Proof it still works Eval suites score every workflow against fixed test cases. Run them after each prompt or model change, and a quality regression is a failed test before promotion, not a customer complaint after. ## Where Your Data Actually Lives The first question every security team asks, answered before they ask it. Orkindo is not a shared multi-tenant SaaS: every customer runs on an instance that is theirs alone. ### A dedicated instance per customer Your own domain, your own PostgreSQL, your own Redis, your own containers, your own network. Nothing is shared with another customer, and the instance size grows from small to large as usage does. - Own domain - Own database - Own network ### Our cloud, your cloud, or your data center Run the same instance managed by us, in your own cloud account, or on premise. The platform ships as containers, so the deployment profile is a choice, not a different product. - Managed - Self-hosted - On-premise ### Reach databases behind your firewall An ERP on a private LAN with no public IP is not a blocker. A secure tunnel configured from System Settings lets agents reach it without opening a single port or exposing the database to the internet. - No public IP - No open ports - Set up in System Settings ### Keep inference inside, when you want to Choose the model per workflow. Run open models locally through Ollama and no prompt, document, or query result ever leaves your instance. Use a hosted provider where accuracy demands it, and know exactly which one sees what. - Local models via Ollama - 100+ providers - Per-workflow choice ### Built to be adopted, not just installed A platform nobody uses is the most expensive kind. Orkindo ships with the mechanics that get a company past the pilot. - **Trilingual product**: The whole application, not just this website, ships in English, Portuguese (Brazil), and Spanish. A team in São Paulo and a team in Madrid use the same instance in their own language. - **Guided onboarding**: A product tour and a getting-started checklist walk each new user from first login to first working agent, without a training session. - **Adoption you can measure**: Levels, badges, streaks, quests, and a leaderboard reward the people who build and use agents. You see who is adopting the platform and who needs a nudge. - **Desktop and mobile**: A desktop application for Windows, macOS, and Linux, and an installable mobile workspace with push notifications for support agents. Same instance, same permissions, every device. ## Built for Every Corporate Need See how enterprises use Orkindo to automate processes with autonomous agent workflows, from scheduled reports to event-driven alerts to on-demand chat access. ### Automated Reporting Workflows Scheduled autonomous agents that generate KPIs, sales summaries, and operational reports and deliver them via WhatsApp or Google Chat every morning. The plant manager reads yesterday's production numbers with the first coffee, and nobody had to ask. Key benefits: - Automatic scheduling - WhatsApp delivery - Zero manual intervention ### Smart Alerts and Monitoring Autonomous agents that monitor inventory, production metrics, and critical indicators in the background. When stock crosses its threshold, the right person hears about it in minutes, not in next week's report. Key benefits: - Continuous monitoring - Automatic alerts - Proactive action ### Conversation With Your Data Managers chat with an agent on WhatsApp or Google Chat to ask about sales, inventory, and KPIs in plain English. Answers come back as tables and charts straight from the database, with follow-up questions that keep the context, no BI ticket required. Key benefits: - Plain-English queries - Tables and charts - Instant answers ### Knowledge Retrieval Workflow An agent workflow that surfaces answers from manuals, policies, and procedures. Triggered automatically for new hires or on demand by employees over chat or WhatsApp, with citations pointing back to the original document. Key benefits: - Cited document answers - Source links - Faster onboarding ### Field Operations Workflow An agent workflow for field technicians: procedures, equipment manuals, and maintenance history available on demand via WhatsApp. Instant answers on-site, no need to return to the office. Key benefits: - WhatsApp access - On-site answers - 24/7 availability ### Branded Internal Apps Ship a workflow with its own UI: an intake form for the support team, an approval queue for finance, a multi-step configuration wizard for ops. Embed it on any internal page, no separate frontend project, same governance as the workflow. Key benefits: - Forms and dashboards - Approval queues - Embed anywhere ## The Brain at Work, in Six Stories Six industries, one arc: work that was drowning a team, an agent that took it over, and people back on the decisions that actually need them. ### E-Commerce: Round-the-clock support that scales without new hires Company profile: Online retailer with 50,000 monthly orders The Challenge: Picture the support inbox at midnight: 2,000+ tickets a day, most of them "Where is my order?" and "How do I return this?", wait times creeping up, and a team working overtime just to stand still. The Solution: Now a WhatsApp agent picks up first. It reads the live order database, checks the returns policy in the knowledge base, and answers on the spot. When a case gets genuinely hard, it hands the thread to a human, and the customer never notices the switch. The Result: Most tickets never reach a person. The ones that do are the ones that should, and the team finally handles them without a queue on fire behind it. - WhatsApp channel - Live database answers - Human takeover ### Manufacturing: Production answers in seconds, no analyst queue Company profile: Mid-size manufacturer with complex production data The Challenge: Three analysts stood between an entire company and its own production data. Every simple question became a ticket, every ticket a wait of days, and every wait a decision made late. The Solution: Today a plant manager types "How many units shipped yesterday?" into WhatsApp on the factory floor and has the number before the coffee cools. Every morning, a scheduled workflow delivers the daily summary before anyone thinks to ask. The Result: Answers in seconds instead of days, analysts doing analysis instead of fetching numbers, and no extra hiring as the company grew. - Plain-language questions - Scheduled summaries - Mobile access ### Financial Services: Always-current answers with a full audit trail Company profile: Regional bank with strict compliance requirements The Challenge: The rules lived in hundreds of PDFs that changed constantly. Employees quoted last quarter's policy without knowing it, and every outdated answer was a regulatory incident waiting to happen. The Solution: Now an employee asks the question in Google Chat and the agent answers from the latest indexed version of the policy, citation attached. Behind the scenes, every single query lands in the audit log, exactly where the regulators want it. The Result: Lookups that took an afternoon take seconds, nobody acts on a superseded policy anymore, and the audit trail writes itself. - Google Chat channel - Source-cited answers - Audit trail ### Education: Personalized help around the clock, a human tutor one step away Company profile: Online course provider supporting many learners at once The Challenge: A student stuck at two in the morning had two options: give up or wait for daylight. Tutors, meanwhile, spent their days re-answering the same routine questions instead of teaching. The Solution: The chat bubble on the site and the WhatsApp agent answer from the course material, check the student's own progress records to tailor the reply, and when someone is still stuck, hand them to a human tutor with the whole conversation attached. The Result: Every student gets help at their own pace at any hour, tutors get to teach again, and the struggling ones reach a person before they quit. - Website widget - Personalized answers - Tutor handoff ### Insurance: Claims triaged in minutes, routed to the right adjuster Company profile: Insurer handling a steady volume of claims across personal and commercial lines The Challenge: A claim arrived by whatever channel the customer preferred, then sat in one undifferentiated queue. Small claims waited behind complex commercial losses, and under that volume, odd patterns slid through unnoticed. The Solution: Now the agent meets every claim at the door, on WhatsApp, web form, or email. It answers coverage questions straight from the policy wording, fast-tracks the simple claims, routes the complex ones to the right adjuster with a summary, and holds anything unusual for a specialist before payout. The Result: First response in minutes on every claim, adjusters spending judgment where judgment is needed, and suspicious claims caught instead of paid. - Multi-channel intake - Automatic triage - Specialist review flags ### Marketing: A standing team of agents ships the weekly content Company profile: A marketing team with more content ideas than hands to execute them The Challenge: The content calendar was a treadmill: research, draft, SEO pass, review, repeat, with a human chasing every hand-off. The week disappeared into coordination, and the calendar slipped anyway. The Solution: So the team hired agents. A Content Lead, a Researcher, a Copywriter, and an SEO Specialist pick up tasks from a shared board on a weekly schedule, pass the work down the line, and send one finished piece to a human for a single go or no-go. The Result: The calendar keeps itself now. People make one approval decision a week instead of fifty coordination ones, and the team budget caps what the agents can ever spend. - Agent Organization - Approval gate - Team budget ## Enterprise Workflow Management at Scale The numbers behind the platform: model providers, database engines, channels, and the library you start from. - **100+** LLM Providers — Supported out of the box - **13** Database Engines — PostgreSQL, Oracle, BigQuery, Snowflake... - **15** Channels — WhatsApp, Slack, Teams, voice... - **222** Ready-to-Run Assets — Templates, workflows, and tools - **3** Product Languages — English, Portuguese, Spanish - **99.9%** Uptime SLA — Enterprise reliability Security & compliance: - AES-256 Encryption - Built to SOC 2 principles - GDPR-aligned controls - Workflow-Level RBAC ## Frequently Asked Questions ### How long until our first agent is in production? Up to two weeks, and usually one. We start from the process that hurts most, connect the data sources and channels it needs, build the agent on a working template, and put it in production with monitoring and approval gates in place. From there, every following agent reuses the same connections and ships faster. ### What is Orkindo built for? Orkindo is built for running autonomous AI agent workflows, agents that execute on their own on schedules or when events occur (daily reports, stock alerts, order processing, monitoring, and more) without human intervention. The same agents are reachable on demand over WhatsApp, Slack, Microsoft Teams, Google Chat, web chat, and 15 channels in total, and when the work is a whole project rather than a single task, Agent Organizations let a standing team of agents carry it from plan to approved deliverable. ### How do autonomous agent workflows work? Agent workflows are configured to execute tasks automatically on schedules (e.g., every day at 8 AM) or when events occur (e.g., new order received). They run in the background without human intervention, generating reports, sending alerts, processing data, holding natural-language conversations with your databases and documents, and reasoning over the results. ### Do I need coding experience to build workflows? Basic Python knowledge is helpful, but our AI copilot makes it easy for anyone to get started. It learns from the workflows, tools, databases, and knowledge bases already on your platform, so its suggestions get better with every workflow you create. You can test instantly with the integrated chat on the same page. Pre-built templates also help you launch workflows in minutes. ### Can employees access these agents directly? Yes. The same agents that run autonomously can also be reached on demand over WhatsApp, Telegram, SMS, Messenger, Instagram, LINE, email, Google Chat, the website chat bubble, or a realtime voice call in the browser. Employees ask in plain English and the agent has a real conversation with your databases and documents, returning tables, charts, and source-cited answers on the spot. ### How does Orkindo handle workflow permissions? Orkindo provides granular, role-based access control at every level. You control who can create and edit workflows, who can trigger or interact with each agent, and which data sources and tools each agent can access. Full audit logs track every action. ### How do I monitor and control LLM costs? LLM Tracing shows every model call your workflows make, broken down by workflow and by model, with token counts, cost estimates, and duration over any time window. Eval Suites let you regression-test workflows after a prompt or model change so you catch quality drops before they ship, and per-role token usage reports help you allocate spend across teams. ### Which AI models does Orkindo support? Orkindo supports 100+ LLM providers including OpenAI, Anthropic, Google, Azure, AWS Bedrock, and local models via Ollama. You can assign different models to different workflows and switch providers without changing agent logic. ### Can Orkindo integrate with our corporate systems? Yes. Orkindo connects to PostgreSQL, MySQL, BigQuery, Oracle, SQL Server, Snowflake, and 13 database engines in total, plus knowledge collections, REST APIs, webhooks, and enterprise tools. Your own systems talk to agents through the API channel, autonomous workflows trigger and interact with anything your business uses, and the same agents stay reachable on demand for a plain-English conversation with the data behind them. ### Do you offer a trial or proof of concept? Yes! We offer no-obligation trials for qualified businesses. During the trial, you can build autonomous agent workflows, schedule them, connect your data sources, expose them on WhatsApp and Google Chat for on-demand access, and evaluate the full platform with dedicated support. ### Can I build internal tools and dashboards on Orkindo? Yes. Any workflow can ship with its own branded UI: intake forms, dashboards, approval queues, multi-step wizards. Embed it on any internal page or run it standalone, no separate frontend project, no extra hosting, and the same versioning, audit trail, and access controls as the workflow it belongs to. ### Can workflows be combined into larger pipelines? Yes. Workflow orchestration chains your existing workflows into one coordinated run: draw the pipeline on a visual canvas, connect the steps so one output feeds the next, and let independent steps run in parallel. Schedule the whole pipeline on a cron, start it from an external system with one webhook call, and follow every run node by node down to each step's execution log. ### What are Agent Organizations? A standing team of AI agents arranged in an org chart: a lead and specialists, each backed by a real workflow underneath. Work flows through a shared task board, on demand or on a schedule, stops at the human approval gates you place, and stays within team and per-agent budget caps. Recurring projects like the weekly content calendar run themselves; you make the final call. ### Can human agents and AI share customer support? Yes. Attendance turns any channel, WhatsApp, Telegram, Instagram, email, the website widget, and more, into a support desk with shared inboxes. Run it bot-first so the AI handles the routine volume and a human takes over any conversation in one click, or route selected channels to human agents only, with round-robin assignment, labels, and canned responses built in. ### Where does our data live? On an instance that is yours alone: own domain, own PostgreSQL, own Redis, own containers, own network. Nothing is shared with another customer. That instance can run managed by us, in your own cloud account, or on premise in your data center. If a workflow uses local models through Ollama, prompts and documents never leave the instance at all. ### Can an agent modify our production database? No. For every database you register, the platform provisions a SELECT-only role and connects agents through it, never through the admin credential. For object storage it does the same with a list-and-get-only user. An agent can answer any question about the data and cannot change a single row. Actions that should write, such as creating a ticket, go through explicit tools you grant on purpose. ### Our databases are on a private network. Can Orkindo reach them? Yes. A secure tunnel configured from System Settings lets agents reach a database on a private LAN with no public IP and no open inbound ports. The database is never exposed to the internet, and the connection is tested before it can be saved. ### Are we locked in? Your workflows are Python files, your tools and templates are files, and the orkindo CLI pulls all of them to a local folder you can version in git. Models are interchangeable across 100+ providers without changing agent logic. What you build is yours to read, export, and move. ### Our developers already use Claude Code. How does that fit? Two ways. Claude Code, Codex CLI, and OpenCode can run inside the platform, with session continuity, workspace sync to a storage connection you control, and every session in the audit log, so IT gets governance without taking the tools away. And Claude Code or Cursor on a developer's machine can connect to your Orkindo workspace over MCP to read and write workflows, tools, and knowledge collections from the terminal. ### Can we add our own capabilities to agents? Yes, at three levels. Skills are reusable instruction packs. Plugins bundle Skills with the MCP servers they need. Marketplaces are Git-backed catalogs you sync and install from, including the official Claude Code and Codex marketplaces or your own private repository of approved packs. Administrators decide which sources each workspace may pull from. ### Do we still need a separate BI tool? Metabase ships with the platform, and every database you register appears in it automatically, so dashboards and the chat answers draw on the same governed connections. Internal Apps can also turn a workflow's output into a live dashboard embedded on any internal page. ### Which languages does the product support? The whole application ships in English, Portuguese (Brazil), and Spanish, and each user picks their own language on the same instance. Agents themselves answer in whatever language the customer or employee writes in. ## Blog ### We already use Claude Code and Codex. Why would we need Orkindo? Published: 2026-10-05 The model is the same. What changes is where the agent lives, who starts the work, who can use it, and who answers for it. An honest comparison between using the models directly and running agents on a platform. Read: https://orkindo.com/en/blog/orkindo-vs-direct-models ### Field notes: when the agent costs more than the developer Published: 2026-07-05 A single agentic task fires off 5 to 30 model calls. Without per-task tracking, cost becomes a number you only discover on the invoice. Read: https://orkindo.com/en/blog/agent-cost-engineering ### Choosing the Right Model for Every Workflow Step: Accuracy Without Overspending Published: 2026-05-21 Using the most powerful model everywhere feels safe. It is not. Here is how to match model capability to task complexity, cut costs significantly, and often improve accuracy at the same time. Read: https://orkindo.com/en/blog/choosing-the-right-model ### A Field Guide to Open-Source Python Agent Harnesses Published: 2026-05-20 Eight open-source Python harnesses for building AI agents, from deepagents and LangGraph to the Claude Agent SDK. What each one is good at, where it struggles, and how to choose between them. Read: https://orkindo.com/en/blog/open-source-python-harnesses ### Multi-Agent Systems Patterns: A Practical Guide to Designing Agent Workflows Published: 2026-05-03 One agent is rarely enough for real enterprise work. The question becomes how to coordinate many of them. Here are the multi-agent patterns that hold up in production, and how to pick between them. Read: https://orkindo.com/en/blog/multi-agent-patterns ### Shadow AI: The Hidden Risk Lurking in Your Organization Published: 2026-04-12 When employees adopt AI tools without IT oversight, they open the door to data leaks, compliance violations, and security blind spots. Here is what Shadow AI looks like and how to get ahead of it. Read: https://orkindo.com/en/blog/shadow-ai ### Welcome to the Orkindo blog Published: 2026-04-10 Product updates, engineering notes, and stories from the field as we build the enterprise AI agent platform. Read: https://orkindo.com/en/blog/welcome ## Contact Let's discuss how Orkindo can replace your repetitive processes with autonomous agent workflows, and give your teams direct chat access when they need it. Grab a time with our team, or reach us by email. - Book a call: https://m2soft.orkindo.ai/w/orkindo-book-call - Email: sales@orkindo.com - Website: https://orkindo.com What to expect: - Response within 24 hours - Personalized demo of relevant features - Custom pricing based on your needs - No obligation trial available - First agent in production in up to two weeks, usually in one --- Autonomous AI agent workflows, teams of AI agents, and a human plus AI support desk for the enterprise, all in a single platform with enterprise governance.