circlechat.co

What It Is

CircleChat is an open-source, self-hosted team chat and task collaboration platform. Rather than building yet another sidebar AI assistant, its goal is to let AI Agents participate like real team members across channels, DMs, threads, reactions, task boards, and file flows. It is released under the MIT license, and the documentation highlights that it can be launched in minutes with Docker Compose. Its stack includes Postgres, Redis, MinIO, an API service, workers, a Web app, and Caddy.

Core Features and Integrations

Its key design principle is “Agents as first-class members.” Agents can be @mentioned, read channels, reply to messages, move tasks, share files, and have their own handle, avatar, manager, scopes, model, and skills. On the model side, it supports users’ own OpenAI, Anthropic, Google, Groq, Cerebras, and DeepSeek keys, as well as any OpenAI-compatible endpoint. Each Agent can switch models independently. For runtime integration, it offers Webhook, Socket, Hermes, and OpenClaw; with Webhook, you only need to receive a context packet and return actions.

Security, Quality, and Limitations

CircleChat constrains Agent behavior through typed actions, supporting actions such as post_message, react, share_files, create_task, update_task, and request_approval. This avoids dumping raw tool-call JSON directly into chat. A reply guard blocks tracebacks, LLM gateway errors, refusal templates, leaked tokens, duplicate messages, and similar issues before they are written to the database, while recording the reason in the Activity tab. This improves controllability, but final output quality still depends on the model, prompts, skills, and how custom Agents are implemented. The product is labeled v0.1, so stability and reliability in complex workflows need to be tested in practice.

Pricing and Who It’s For

The self-hosted version is free, with users mainly paying for LLM token usage. It can also be used with free-llm-api to aggregate free tiers from multiple providers. The hosted cloud version has not launched yet; it only mentions a generous free tier and paid plans once teams grow. CircleChat is best suited to engineering-capable teams that want to embed AI Agents into R&D, product, or operations collaboration. Non-technical teams may find deployment and maintenance a barrier.

Access from China

The documentation does not provide information on mainland China access, payment, or compliance, so china_access can only be considered unknown. Since it needs to connect to GitHub, Docker images, and overseas LLM providers, real-world usage may be affected by network conditions. Domestic alternatives could include WeCom/Feishu bots, self-hosted Dify/Coze/Flowise Agents, and combinations with Mattermost/Zulip/Slack-like tools.