ComposioHQ/trustclaw — A self-hostable personal AI agent with vector memory, Composio tools, and Telegram.

Description and use

**Brief Summary**
ComposioHQ / trustclaw is a self‑hosted personal AI agent built in TypeScript that combines vector‑based memory, Composio’s suite of productivity tools, and Telegram integration. It lets developers prototype Retrieval‑Augmented Generation (RAG) or autonomous agent workflows without assembling a model stack from scratch, and it ships with ready‑to‑use tooling for rapid experimentation.

**Value**

• **Speed to prototype** – All the heavy‑lifting (vector store, tool wrappers, messaging channel) is pre‑wired, so teams can focus on the business logic of their AI feature rather than on infrastructure.
• **Modular toolset** – By exposing Composio’s APIs, trustclaw makes it easy to add calendar, email, or database actions to an agent, turning a simple chatbot into a functional personal assistant.
• **Open‑source transparency** – With ~350 stars and an active TypeScript codebase, the project is easy to audit, extend, and integrate with existing TypeScript/Node.js stacks.

**Practical Adoption Path**

1. **Clone & spin‑up** – Fork the repo, run the Docker compose (or npm install + npm start) to launch the agent locally.
2. **Configure secrets** – Add your OpenAI (or other LLM) API key, vector‑store credentials, and Telegram bot token in the .env file.
3. **Define tools** – Use the provided Composio SDK to register the specific APIs your use case needs (e.g., CRM, calendar).
4. **Iterate** – Interact via Telegram, observe the agent’s memory snapshots, and refine prompts or tool bindings.
5. **Hardening for production** – Replace the default SQLite vector store with a managed Pinecone/Weaviate instance, enable TLS, and add rate‑limiting/auth checks around the Telegram webhook.

**Production Readiness**

• **Maturity** – Medium. The project is functional for prototypes and internal workflows, but it lacks comprehensive CI/CD pipelines, extensive automated tests, and detailed deployment documentation.
• **Dependencies** – Relies on external LLM APIs, a vector store, and Telegram; each must be evaluated for latency, cost, and compliance.
• **Maintenance** – Recent commit (2026‑05‑13) shows active development, but the core maintainer count is low; organizations should plan for in‑house ownership or a service‑level agreement.
• **Risk considerations** – No immediate licensing or security red flags, but a formal security audit (especially around the Telegram webhook and secret handling) is recommended before exposing the agent to production traffic.

In short, trustclaw offers a fast, open‑source foundation for building personal AI agents, suitable for proof‑of‑concepts and internal tools, while requiring additional hardening and operational oversight for production deployments.

Problem

ComposioHQ/trustclaw helps add AI capability without starting from a blank model stack.

Outcome

Medium: useful for prototypes or internal workflows, with dependency and maintenance checks before production.

Workflow

• prototype AI features
• build RAG or agent workflows
• evaluate model tooling

Evidence

A self-hostable personal AI agent with vector memory, Composio tools, and Telegram.

Source