Comp AI CRM Review

Comp AI CRM is an open-source CRM with an AI agent that reads email, creates contact and company records from new people on threads, enriches profiles with domain, industry, and logo, schedules its own follow-ups, and explains its timing. Standard CRM views (pipeline, company table, deals, activity feed) sit on top. Each record has an Agent tab showing what the agent did and letting reps answer its questions, and the project's rule is that nothing about a person is guessed: weak evidence becomes a suggestion for a human to settle.

Locally you need Bun, Docker, and Postgres (provided by the included docker compose file), then copy .env.example, set BETTER_AUTH_SECRET and ALLOWED_SIGN_IN, add a Google or Microsoft OAuth client, and run bun install, docker compose up -d, bun run db:deploy, and bun run dev. Production is three deployments (the Next.js app, the NestJS API, and the agent) plus Postgres, sharing DATABASE_URL and BETTER_AUTH_SECRET, and you schedule a mailbox sync endpoint with a cron secret. The agent's model runs through Vercel AI Gateway, and the README does not say what credential a non-Vercel install needs.

Day to day it looks like a conventional CRM, with pipeline and table views and URL-based filters, and the novelty is the per-record agent log. We did not deploy it ourselves; this section reflects the repository README and the OpenAlternative listing.

• Contacts and companies created automatically from email threads
• Agent with a work queue, tools, and a sandbox without network access or database credentials
• Agent builder that creates new agents from plain-language descriptions
• Per-record Q&A that shows its reasoning, and a Slack integration for deal events
• Read-only mail sync for Google and Microsoft accounts

• Permissive MIT license, rare among CRMs with AI at the center
• Cautious design: weak evidence becomes a suggestion instead of a guessed fact
• Fast growth, with about 11,000 GitHub stars about two months after launch

• Single tenant by design, with no organizations
• No hosted offering is described, so you must run three deployments plus Postgres
• Web research and company data need extra keys (Perplexity, a Context key)
• Young project: the last commit was about a month before this review, and expired Microsoft client secrets can break mail sync silently

The software is free under the MIT license. Your costs are hosting, Postgres, and the LLM calls behind the agent, plus optional paid data services such as Perplexity.

HubSpot and Salesforce are full platforms with marketing, service, reporting, and large ecosystems. Comp AI CRM covers the core of a sales CRM, contacts, companies, deals, and activity, and makes automatic logging from email its main feature. It is a better fit for a small technical team that wants to own the code, and a weaker fit for teams that depend on marketing automation or extensive integrations.

• Small technical teams that want an AI-first CRM they can fork and change
• Founders who live in their inbox and want contact records created automatically
• Projects that need a permissive license for the CRM code

• Sales teams without engineering support to deploy and maintain it
• Multi-tenant or agency setups
• Teams that need marketing automation, quoting, or a large integration catalog