AI coworkers.
Ready for the enterprise.
OpenBot is the modular, self-hostable, enterprise-ready agent workspace with a clean architecture and built-in governance. Use it as-is, white-label it, or deeply customize it. Bring the agents your engineers build or start with ours, then choose the models, tools, and infrastructure they run on.
Fully self-contained. Free for personal use.
Deploys toAWS, Azure, Google Cloud, or Kubernetes
Enterprise controls built in.
Your infrastructure.
Made by CopilotKit
Trusted in production by the majority of the F500
Give it a task.
Follow the work.
See the work and act on the result. Review findings, work with interactive results, and approve the next step in the same conversation.
Read the misconnect playbook
Read 12 messages in #ops-daily
Checked 212 connecting passengersThree turns ran late, all on the evening bank. KT 482 to Boise is the worst at 52 minutes, and 14 of its passengers would miss their connection. I drafted their rebookings on KT 488 for the desk to confirm, and posted the summary to #ops-daily.
14 rebookings drafted on KT 488
Filed OPS-318 for the Denver station team
Emailed the Boise station leadPosted to #ops-dailyRead the misconnect playbookRead 12 messages in #ops-dailyChecked 212 connecting passengersThree turns ran late, all on the evening bank. KT 482 to Boise is the worst at 52 minutes, and 14 of its passengers would miss their connection. I drafted their rebookings on KT 488 for the desk to confirm, and posted the summary to #ops-daily.
14 rebookings drafted on KT 488
Filed OPS-318 for the Denver station teamEmailed the Boise station leadPosted to #ops-dailyBring your agents.
Keep your stack.
OpenBot is a composable workspace for your agents, models, and tools. Its modular design lets you switch model providers or connect a new tool while keeping the agents and infrastructure you already use.
Keep the agent you built.
Connect Claude Agent SDK, LangChain, Google ADK, or another AG-UI-compatible agent. Your prompts and tools come with it.
Claude Agent SDK
OpenAI Agents SDK
Choose the model for the work.
Use OpenAI, Claude, Gemini, or Grok with your own subscriptions or API keys, or open models like Kimi, DeepSeek, Qwen, Nemotron, and Mistral. Keep the provider relationships you already have.
OpenAIClaude
Gemini
Grok
Kimi
DeepSeek
Qwen
Nemotron
MistralFrom the team behind
CopilotKit and AG-UI.
OpenBot brings the same open standards to a workspace your company can make its own.
Trusted by the majority of the Fortune 500 in production.
See how
OpenBot compares.
Every product here gives you capable AI coworkers. The differences show up later: whose agents, whose models, whose servers, and whose rules.
Comparing OpenBot with Claude, OpenAI dots, OpenClaw.
 OpenBotFor Business | ClaudeAnthropic | OpenAI dotsOpenAI | OpenClawOpen source | |
|---|---|---|---|---|
| Model lock-in | Yes: Future-proof. Any closed or open model, through your subscriptions or API keys | No: Locked in. Claude only, from Anthropic or the big clouds | No: Locked in. OpenAI’s models only, picked by OpenAI | Yes: Future-proof. Any model, from any provider or locally |
| Agents | Yes: Developer-loved. The agents your engineers already write, in the frameworks & harnesses they use — through AG-UI | No: Developers locked out. Your agents can’t run in it, even on Anthropic’s SDK | No: Developers locked out. Only OpenAI’s agent; yours only as tools | Limited: Built around its own agent. Only coding tools plug in, on its own server |
| Modularity | Yes: Modular by design. Great defaults, or plug in your own skins, sandboxes, MCP hubs, and more | No: Closed product. Only tools and sign-in plug in | No: Closed product. Only tools, a computer, and sign-in plug in | Limited: Plugins around its own agent. Swap anything but its agent |
| Ownership | Yes: Yours. Your infrastructure, your brand, and the MIT code | No: Anthropic’s. Anthropic’s service, closed code | No: OpenAI’s. OpenAI’s cloud, closed code | Yes: Yours. Your machines or servers, and the MIT code |
| Built for the enterprise | Yes: Governance designed in, from CopilotKit (SOC 2 Type II) | Yes: A mature Enterprise plan | Basic: Launched last month, on ChatGPT Enterprise | No: Alpha, by its own rating |
| Enterprise support and forward-deployed engineers | Yes: Full support, plus engineers who embed with your team | Limited: Engineers only for top accounts | Limited: Through a separate consulting arm | No: Community support only |
Connect the tools.
Control the access.
Connect your existing MCP tools or choose integrations from the available catalogs. Give each Bot the access it needs, with company policies governing what it can do.
Choose what each Bot can reach.
Enable the services your team needs, then grant tools to the Bots that need them. People connect their own accounts.
Set boundaries before it acts.
Block actions that should never happen. Require approval for sensitive work, so people keep the decisions that matter.
Know what happened.
Review the audit trail when a task needs attention. Your admins can trace actions and decisions instead of chasing context across chat windows.
Agents that learn
and keep improving.
Teach your agents directly, or let them learn from everyday work. OpenBot observes patterns, develops reusable skills, and improves how tasks get done, with people in control.
Trust is a setting you control.
Decide what agents can do, what needs approval, and what stays off limits. Give your security team a record of the actions taken.
Read ticket#4821
Postedin #supportPosted in #support and assigned #4821 to Billing.
Access check
Posting to #supportAsk a person firstAllowed by Dana Ruiz
Recorded 10:42:07
Teach a process. Let experience improve it.
Show an agent how a task is done to teach it directly. It also learns as it works, picking up patterns and turning successful approaches into skills. You don’t have to teach every step for your agents to get better.
Run it where
you need it.
From a developer’s laptop to your company’s private cloud. Deploy into your own network in one click, or talk to us about a hosted deployment.
Deploy in one click.
Into your own AWS or Azure network, Google Cloud, or any Kubernetes cluster. One Helm chart, running inside your network.
Anthropic
OpenAI
GeminiStart where you work.
OpenBot runs on your laptop as a self-contained workspace. Your workspace data and API keys are stored on your machine. When a Bot uses an external model or tool, it sends that service the data it needs for the task.
- Open-source application you can customize
- Your preferred models and agents
- Set up with your coding agent
Governance built in.
Not bolted on.
OpenBot gives your security, IT, and finance teams the controls they expect. Decide who gets in, what each Bot can do, and what it can spend, with a record of everything it touches.
Identity and access
- Single sign-on
- People sign in through your company’s identity provider, with no separate passwords to manage.
- SCIM provisioning
- Users and groups sync from your directory, so access starts on day one and ends when someone leaves.
- Bot roles and on-behalf-of tokens
- Give Bots custom roles, and let them act for people with access tokens minted on their behalf, usually with a subset of their permissions.
Control
- Admin plane
- Manage people, Bots, tools, and policies for the whole company from one place.
- Tool access at every level
- Grant tools per Bot, per role, and per person, and let people ask an admin when they need more.
- Cost audit and controls
- See and control spend per Bot, per role, per team, and per person.
Visibility
- Full audit trail
- Every item accessed and every action taken is recorded, so you can always see who did what.
- Audit log export
- Send audit events to the SIEM your security team already uses.
Compliance
- SOC 2 Type II
- CopilotKit, the company behind OpenBot, is SOC 2 Type II compliant.
- EU AI Act labeling
- Watermark and label AI-generated content, in line with the EU AI Act’s transparency rules.
- Retention and deletion
- Decide how long conversations and logs are kept, and delete a person’s data on request.
Hosting and tenancy
- Fully self-hostable
- Run OpenBot on your own infrastructure, so stored data stays in the network and region you choose.
- Multi-tenancy
- Isolate each customer’s data in its own tenant, with its own settings and access controls, while super-admins oversee them all.
- Usage limits
- Set daily, weekly, and monthly token limits per tenant and per person, and see spend for every tenant.
Slack · Post messages
- BotSupport BotAllowed
- RoleSalesAsk an admin
- RoleContractorsBlocked
- BotSupport Bot
- RoleSales
- TeamFinance
Work with your Bots.
Wherever work happens.
On iOS, Android, Slack, and Teams, as well as desktop and web. Linked identities keep every conversation and approval connected to you.
DesktopYour Bots, tools, and conversations on your Mac or Windows laptop.
Web Coming soonOpen your workspace from any browser.
Request web accessOne executive.
20 AI coworkers.
One person sets the direction. An orchestrator coordinates specialists across six functions, with a person approving the work that matters.
Read how the team works
This is a real story of a company that built its entire go-to-market team on OpenBot, with 20 AI agents in real roles and one executive setting the direction.
The brief
An executive needed to build a go-to-market team to handle all of the company’s sales and marketing work, except meeting directly with buyers. That part had to be done by people.
Using CopilotKit OpenBot, the company made 20 AI agents, and one person built them all and runs them.
How the team is built
Each agent holds a real GTM role, such as growth strategist, ABM specialist, demand gen manager, enterprise SDR, RevOps engineer or pipeline strategist. Each one has a written job description and works in its own tools. Across the team, those come from seven products:
The agents picked their own names, each after a TV or film character who fits the job. That is how the demand gen manager became Don Draper, the program manager became Leslie Knope and the librarian became Rupert Giles.
One agent sits above the specialists. Nicky Fury, the orchestrator, takes direction from the executive, assigns each piece of work to the right specialist and coordinates the hand-offs between them.
How a cycle runs
Work starts with a message from the executive. It can be a single line, such as “Build pipeline for the AWS launch by Friday.” Nicky Fury works out who should do what and passes each part to a specialist, who does it in their own tools and hands the result on to whoever needs it next.
When the outreach is ready, the executive approves it before it goes out.
Every night, a QA analyst audits the team’s output. A program manager runs the work in two-week GTM cycles, and a librarian keeps the team’s shared memory, so lessons from one cycle carry into the next.
Who does what
The executive sets the strategy and makes the decisions only a person can make, and people still hold every meeting with buyers. The agents do the rest of the work, across strategy, marketing, sales, data and RevOps.