Graph Indexer is a local MCP server that works with ten AI coding agents: Claude Code, Cursor, VS Code Copilot, Codex, Gemini CLI, OpenCode, Kilo Code, Junie, Zed and Devin. Run npx graph-indexer init in your repository and it writes the project-level MCP configuration for each agent it detects, so the agent gets the same eight read-only tools for exact references, call graphs and change impact.
Which agents does Graph Indexer support?
The table lists each agent, the file npx graph-indexer init writes for it, and the page with the exact configuration. Agents marked “printed” have no project-level config file that init can edit, so it prints the entry to paste instead.
| Agent | File written by init | Guide |
|---|---|---|
| Claude Code | .mcp.json (plus .claude/agents/code-structure.md) | Claude Code |
| Cursor | .cursor/mcp.json | Cursor |
| VS Code Copilot | .vscode/mcp.json | VS Code Copilot |
| Codex | printed TOML for ~/.codex/config.toml | Codex |
| Gemini CLI | .gemini/settings.json | Gemini CLI |
| OpenCode | opencode.json | OpenCode |
| Kilo Code | opencode.json (or an existing kilo.json) | Kilo Code |
| Junie | .junie/mcp/mcp.json | Junie |
| Zed | .zed/settings.json | Zed |
| Devin | printed server entry for Devin’s MCP settings | Devin |
What do all the agents get in common?
Every agent talks to the same stdio server and sees the same tools: read_code, search_text, search_code, find_references, call_graph, change_impact, check_changes and outline. The tools reference lists their parameters. Graph Indexer itself makes no network calls and sends no telemetry; the agent still sends what it reads to its model provider.
init also adds a short managed block to CLAUDE.md or AGENTS.md that tells the agent when to use find_references, call_graph, change_impact and check_changes. It merges into existing config files and never overwrites other MCP servers.
What if my agent is not on the list?
Any MCP client can run Graph Indexer over stdio with the command npx -y graph-indexer@3 serve --repo /path/to/repo. init prints the same instruction for Claude Desktop, JetBrains AI Assistant and Cline. See Install Graph Indexer for the manual configuration and the Node.js 22.5+ requirement.
Which agent was the benchmark run on?
The agent benchmarks were run with Claude Code, so the cost and time figures on the benchmarks page describe that agent. The reference-accuracy numbers measure the tools themselves and do not depend on the agent. Results for other agents may differ, and Graph Indexer does not publish measurements for them.