Verified project record

trycompai/crm

trycompai/crm is an open source, agent-first CRM where an autonomous agent researches contacts, records verified facts, and books follow-ups on its own schedule. It can be self-hosted and run locally, with the CRM acting as the place where the agent keeps its notes.

Project overview

It inverts the usual CRM model: instead of an AI feature bolted onto a CRM, the agent does the research and note-taking work itself while the CRM serves as its record store, which is a distinctive take on automating CRM upkeep.

Project type: AI Agent
Deployment: Refer to project documentation
License: MIT

Best for

Developers willing to self-host and set the project up themselves who want CRM research, fact-recording, and follow-up scheduling handled by an autonomous agent.
Teams that are fine with a single-tenant setup and want to watch the agent's steps, discarded leads, and answered questions per record.

Key capabilities

Every contact, company, and deal has an Agent tab showing the agent's steps, discarded leads, and questions answered in place.

Limitations and risks

There are no organizations; the project is single tenant, deliberately.
Model access goes through Vercel AI Gateway, and optional integrations such as the Perplexity API, Context, Google OAuth, and Microsoft OAuth involve external services; whether data leaves your environment depends on how these are configured.
Setup requires coding ability, including working with Bun, Docker, .env configuration, migrations, and OAuth provider setup.

Getting started

Setup is of medium difficulty: it requires Bun, Docker, .env configuration, migrations, and an OAuth provider setup.
Copy .env.example to .env and fill in the values, run docker compose up -d then bun run db:deploy, then run bun run dev and open http://localhost:3000.

Evidence and sources

GitHub project description: Comp AI CRM is an open source, CRM designed for AI agents.
README: Most CRMs are a database with a form in front of it. The AI ones bolt a chat box onto the side of that form. Both leave the actual work — finding out what is true, and writing it…
README: It runs on its own deployment, on its own schedule, against its own work queue. It decides what to look at next, books its own follow-ups, spends a research budget, and stops when…
README: [`apps/agent`](./apps/agent) is its own deployment, built on [**eve**](https://eve.dev) — Vercel's filesystem-first framework for durable agents.
README: | **Model** | [Vercel AI Gateway](https://vercel.com/docs/ai-gateway) — no provider SDK, and OIDC on Vercel means no key to manage |

Source updated at: 2026-10-23T18:09:41Z

Open original GitHub project