Comp AI CRM
An open-source CRM built so AI agents can run research and keep records themselves.trycompai/crm
- Stars
- 11.8k
- Forks
- 1.8k
- Licence
- MIT
- Language
- TypeScript
- Last update
- 4 days ago
- Type
- App
What it is
This is a customer relationship manager (a tool for tracking contacts, companies and deals) that is designed around an AI agent rather than a person clicking through forms. The agent runs on its own schedule, looks up information about contacts and companies, and writes what it can verify into the records, while leaving weak evidence as suggestions for a human to decide. The README says it is built with eve, Bun, Postgres and Vercel.
- Have an agent check your own email threads and meeting notes to fill in contact details
- Let an agent research a company's brand data and logo when an account is added, if a Context key is configured
- Watch the agent's steps and questions in an Agent tab on each contact, company or deal
- Connect external services in Settings → Connections with read-only or read-and-write access per tool
Great for
- Teams that want an AI agent to research and update customer records on a schedule
- Developers who want to self-host a CRM and experiment with agent-driven workflows
- People who want their own thread, meeting and email data used as the main source of contact information
- Organizations that prefer MIT-licensed software they can inspect and modify
Not ideal for
- Non-technical users who need a ready-to-use hosted CRM, since setup involves a monorepo, Bun and a Postgres database (reason: the README is developer-oriented)
- Teams that need proven, mature CRM workflows, since the project's approach is new and the README does not describe its stability (reason: no maturity information is given)
- Anyone who wants the agent to act without oversight, since the design routes weak evidence to a human (reason: the project intentionally requires review)
How to get it
- 1
Clone and build from source
You'll need: Bun · Postgres · Vercel account (for the deployment described in the README)
Clone the repository
git clone https://github.com/trycompai/crm - 2
Run it
- Clone the repository and follow the Quick start section of the README (not included in the excerpt)
- Set up a Postgres database and the required environment variables such as DATABASE_URL
- Optionally set AGENT_BRIDGE_SECRET to the same value in both processes to enable the Agent tab
- Add optional API keys such as PERPLEXITY_API_KEY or a Context key in Settings → General or the environment
- 3
Get the most out of it
- Add a Context key to enable company brand data and LinkedIn lookups, changeable later in Settings → General
- Add a PERPLEXITY_API_KEY to enable web research by the agent
- Use Settings → Connections to review each discovered tool and grant agents only the access they need
- Read docs/agent.md for the full description of the agent design
Downloads
Strengths
- Keeps agent behavior inspectable: steps, discarded leads and reasons are shown on each record
- Works without any outside API keys, using only the workspace's own data
- Access control is explicit: connections, tool reviews and per-agent grants are described in the README
Trade-offs
- Setup requires several technologies (Bun, Postgres, Vercel deployment) and the README's quick start is not included in the excerpt
- Optional features depend on third-party keys such as Perplexity and Context, which add cost and dependencies
- The README is truncated, so details on hosting costs, maturity and full configuration are not verifiable here
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.