OpenBot 1.0.0: an environment for AI agents with a dedicated computer for each bot
Version 1.0.0 of the OpenBot platform was released on 09.11.2026. It is a modular self-hosted environment for AI agents: each bot gets its own container, browser, files and tools, and every action is recorded before and after execution. Versions 0.3.0 through 1.0.0 have shipped since the previous review.
What's new
- Support for multiple accounts of the same app. Migrations 0084 and 0085 add the brokered_accounts table and an account column in approval_rules — you can now connect several accounts of the same service and set approval rules for each one separately.
- Backend updated to CopilotKit Intelligence 0.16.0-rc.10. OpenBot Desktop updates its backend on the next launch, and data is preserved.
- Downloads via downloads.copilotkit.ai. OpenBot Desktop and its container engine contact downloads.copilotkit.ai first and fall back to GitHub, ghcr.io and Docker Hub only if it is unavailable. This does not apply to self-hosted Compose and Helm.
- Approval of skills learned from recordings. Migration 0083 creates the learning_skill_approvals table, which records who approved a skill and whether access to it was requested.
- Renaming of goal tools. create_responsibility and seven other tools are renamed to create_goal and similar names; the old names are removed, so bot instructions and skills need to be updated.
- Confirmation for archive and destructive actions. App tools with the archive action or marked as destructive now always ask Allow once or Deny, regardless of settings.
What the platform does
OpenBot gives each bot its own container with a /workspace working volume and a separate browser profile; where the host supports it, they can run under gVisor via COMPUTER_RUNTIME=runsc. The platform supports MCP: the catalog includes Google Drive, Notion and the research providers Parallel, Exa and Firecrawl, and Composio connects several hundred more apps. Conversations survive restarts through CopilotKit Intelligence, and each deploy tags its own threads.
Your own model instead of the cloud
Your own model is connected through an OpenAI-compatible API: the OPENAI_BASE_URL variable points to the endpoint the key is spent against. The platform is installed via npx --yes copilotkit@latest login and npx --yes copilotkit@latest project select. License — MIT.
How it differs from its neighbors
According to the feature comparison, OpenBot has no capability claimed only for it: custom tools and functions, voice and speech, its own server and Docker, and multiple agents working together are mentioned for most neighbors but are not stated in OpenBot's documentation. At the same time, OpenBot stands out with its "computer per bot" approach — a separate container, volume and browser profile for each agent, which the comparison does not note for its neighbors.
| OpenBot | OpenClaw | Hermes Agent | DeerFlow | |
|---|---|---|---|---|
| Type | environment | environment | environment | environment |
| Language | TypeScript | TypeScript | Python | Python |
| License | MIT | MIT | MIT | MIT |
| GitHub stars | 10 587 | 391 715 | 254 468 | 84 219 |
| Releases in 90 days | 25 | 18 | 20 | 1 |
| Multiple agents working together | — | — | yes | yes |
| Custom tools and functions | — | yes | yes | — |
| MCP support | yes | — | yes | — |
| Agent memory | yes | — | yes | yes |
| Knowledge base (RAG) | — | — | — | yes |
| Browser control | yes | — | — | yes |
| Human approval | — | — | yes | — |
| Tracing and debugging | — | — | — | yes |
| Voice and speech | — | yes | yes | — |
| Own server and Docker | — | yes | yes | yes |
| Own model: Ollama | — | yes | yes | — |
| Own model: llama.cpp | — | — | yes | — |
| Own model: vLLM | — | — | yes | yes |
| Own model: MLX | — | yes | — | — |
| Own model: OpenAI-compatible API | yes | yes | — | yes |
Comparison based on project READMEs and documentation as of 10.11.2026: "yes" means the capability is stated there, a dash means it is not mentioned (which does not mean the capability is absent). Stars and releases are per GitHub data.
It is worth a look for those building agents in a self-hosted environment with approval management and auditing, especially if isolation of each bot in its own container matters. You can follow new releases in the CPU3D news feed.
