ppl - Personal CRM for AI Agents

ppl

Relational memory for you and your AI.

ppl is a personal CRM for you and the AI assistants you choose. Keep contacts, conversation notes, birthdays and follow-ups together. Work in the browser or authorize an assistant to use the same records.

Read about ppl, choosing a personal CRM, or importing your contacts.

The guides and free tools cover client setup, product comparisons and contact migration. Start with your client’s Claude, ChatGPT, Cursor, or VS Code guide. Before importing, the local CSV checker can flag common file issues without uploading your contacts. Writers and directory maintainers can use the press kit for current product facts, the original logo and public screenshots.

Developer home: ppl-crm.github.io. Try the fictional relationship-memory demo without connecting an account.

Connect your AI

Hosted MCP server: https://withppl.com/mcp

The hosted service is available now. A self-hosted version is coming soon. Your agent connects over Streamable HTTP with a Bearer token.

Start with your client

Use the connection guide for client setup or the developer guide for authentication, permissions, starter recipes and recovery. New users can create an account during connection and verify email ownership. Some clients require a separate configuration step; available features can depend on account plans and administrator settings.

Codex CLI

Codex supports Streamable HTTP MCP servers with OAuth and Dynamic Client Registration. Add ppl, then sign in:

codex mcp add ppl --url https://withppl.com/mcp

codex mcp login ppl

Or add it to ~/.codex/config.toml and run codex mcp login ppl:

[mcp_servers.ppl]

url = "https://withppl.com/mcp"

The login opens ppl in your browser to approve access. New to ppl? Create a free account (free beta, up to 1,000 contacts). Run codex mcp list to confirm the server, or /mcp inside the Codex TUI. See the Codex MCP docs.

Agent connection flow

1. POST https://withppl.com/api/agent/connect/request with {"client_name": "Your agent"}
2. Show the human the approve_url. They sign in or create an account, verify email ownership and review the requested permissions.
3. Poll the returned poll_url conservatively until approved and save the one-time api_token in a protected credential store. Stop on denial or expiration.
4. Use the token with the MCP server or REST API.

MCP clients that support OAuth (Claude, ChatGPT, and others) can connect with standard OAuth 2.0: dynamic client registration, PKCE, and refresh tokens. See docs/oauth.md.

Machine-readable: ppl.json, llms.txt, and the OpenAPI spec for the REST API.

MCP registry

Registered as io.github.ppl-crm/ppl in the official MCP registry. See server.json.

Examples

• Python briefing examples - a dependency-free REST client and an MCP SDK client
• Node MCP client - connect with the MCP TypeScript SDK
• ppl-memory PyPI package - LangGraph, CrewAI, and AutoGen adapters

Docs

• Agent integration guide - task-scoped workflows, tools, and patterns
• OAuth for MCP clients - Dynamic Client Registration flow

License

MIT. See LICENSE.