# Tandem documentation

Tandem is a CRM that your AI agent works in. Your team uses the web app; your agent (Claude, ChatGPT, Codex, Cursor, Hermes, Muse, Delegated or anything else that speaks the Model Context Protocol) uses Tandem's **MCP server**. Both work on the same records, and every change records who made it.

**Tandem's MCP server:** `https://api.tandemcrm.app/mcp`

It speaks MCP over streamable HTTP, signs agents in with OAuth 2.1 (or an API key), and exposes 28 tools for companies, people, deals, activity, tasks, buying signals, suggestions, custom fields and a do-not-contact list. In hosts that support MCP Apps, results appear as interactive views: the pipeline, record cards, task lists and the change feed.

## What an agent can do in Tandem

- **Keep records clean.** Add companies and people without duplicates: companies match on their domain, people on their email.
- **Work the pipeline.** Create deals, move them between stages (lead, qualified, proposal, negotiation, won, lost) and record who the champion and decision maker are.
- **Log activity.** Notes, emails, calls and meetings on any company, person or deal.
- **Book follow-ups.** Tasks with due dates and owners, checked against existing tasks so nothing is booked twice.
- **Find buying signals.** Hiring, funding, job changes, new tech, complaints and more, ranked by timing and strength.
- **Enrich leads.** Fill logos, descriptions and social links from a company's own website, and missing details with the agent's own tools (Apollo, Clay, web search).
- **Ask before acting.** When unsure, the agent proposes a change with its evidence, and a person approves it.

## Where to start

- [Quickstart](/docs/quickstart): connect an agent and run your first prompts in five minutes.
- [Connect your agent](/docs/connect): step-by-step guides for each AI app.
- [Tools reference](/docs/tools): every tool, its parameters and examples.
- [Permissions and scopes](/docs/permissions): what read, suggest, write and delete allow.
- [REST API](/docs/rest-api): the same operations over HTTP, for scripts and integrations.

## For language models

This documentation is also available as plain Markdown: [/llms.txt](/llms.txt) is an index and [/llms-full.txt](/llms-full.txt) is every page in one file. Add `.md` to any docs URL for that page's Markdown.
