Most CRMs were designed for humans clicking through dashboards. OpenClaw was designed for AI agents making API calls.
The new open-source project is a self-hosted CRM built on the assumption that an AI agent - not a person - will be doing most of the data entry, contact management, and deal tracking. It ships with 40+ REST API endpoints, machine-readable documentation, and a skill file generator that lets you drop CRM access directly into an AI agent's configuration.
How the Agent Integration Works
OpenClaw exposes its entire CRM through a REST API with bearer token authentication. The clever bit: it auto-generates a skill file from its OpenAPI spec that you can feed directly to an agent. Eight read-only tools (searching, listing, retrieving records) run automatically. Five write tools (creating, updating, deleting) require human confirmation before executing. That's a sensible split - you want your agent browsing contacts freely but not deleting your biggest client's record without asking.
The built-in chat agent supports Claude, GPT-4o, Llama, and Gemini through OpenRouter, with token-by-token streaming and up to 10 rounds of tool calling per message.
The CRM Itself
Underneath the AI layer, it's a capable CRM: Kanban deal pipelines, 17 attribute types, custom objects, full-text search, CSV import/export, and compound filtering. The tech stack is Next.js 15, PostgreSQL 16, Drizzle ORM, and shadcn/ui - all mainstream choices that should make contributing straightforward.
It's MIT-licensed and self-hosted, which means no per-seat pricing and your data stays on your infrastructure. A live demo is available for anyone who wants to try before setting up PostgreSQL.
The project is early, and "CRM for AI agents" is a bet on a workflow that most teams haven't adopted yet. But if you're already using coding agents and want them to manage your pipeline, this is one of the first purpose-built options.