Supermodel: Code Graph API for AI Agents

# Supermodel Supermodel builds persistent world models for coding agents. The API creates a structural representation of your codebase — functions, classes, modules, call chains, and dependencies — that agents can query, reason over, and return to across sessions and compactions. Code graphs are the persistent world model that gives coding agents architectural understanding of the codebase, always up to date as code changes. ## Product - [Supermodel API](https://supermodeltools.com/#pricing): REST API for querying code graphs — callers, callees, dead code, type hierarchies, module dependencies - [Docs](https://docs.supermodeltools.com) - [Pricing](https://supermodeltools.com/#pricing) ## Open Source Tools (powered by the Supermodel API) - **[Uncompact](https://github.com/supermodeltools/uncompact)** — Claude Code hook powered by Supermodel's code graph API. When Claude Code auto-compacts conversations, agents lose architectural context. Uncompact maintains a persistent world model of your codebase that survives compaction — reloading the structural graph instantly so agents keep working like nothing happened. - **[Supermodel MCP Server](https://github.com/supermodeltools/mcp)** — Brings the full code graph API into AI-native editors (Claude Code, Cursor, Windsurf). Query any symbol, trace any dependency, understand any codebase directly from your agent. - **[Deadcode Hunter](https://github.com/supermodeltools/dead-code-hunter)** — Dead code bloats your codebase and pollutes agent context. Deadcode Hunter uses Supermodel's call graph to flag unreachable functions and unused symbols directly in your pull requests, as a GitHub Action. - **[Test Coverage Map](https://docs.supermodeltools.com/test-coverage-map)** — Maps every test file to the exact functions it exercises using Supermodel's call graph, surfacing untested code paths and showing you what breaks if a test disappears. - **[IDE Extension](https://github.com/supermodeltools)** — VS Code e