Tokennuke
Index local repos so your coding agent can search symbols, graphs, and definitions fast across many languages.
Overview
Tokennuke is a MCP server for the Build phase that indexes your codebase and exposes hybrid search, call graphs, and fast symbol retrieval to your AI client.
What is this MCP server?
- 15 MCP tools for code navigation, search, and graph-style exploration
- Indexes 10 programming languages with hybrid search and call graphs
- Advertises O(1) retrieval for hot paths after indexing
- stdio PyPI package (tokennuke) for local Claude Code / Cursor MCP config
- Open-source repo on GitHub for self-hosted indexing
- 15 MCP tools documented in the server description
- 10 programming languages supported for indexing
- Version 1.3.0 on PyPI with stdio transport
What problem does it solve?
Agents waste context and time re-scanning large, multi-language repos because they lack a persistent structural index.
Who is it for?
Indie builders running Claude Code or Cursor on sizable local repos who want graph-aware navigation without shipping code to a third-party index.
Skip if: Teams that only need occasional grep, cannot run a local Python MCP process, or want cloud-only code intelligence with zero local setup.
What do I get? / Deliverables
After you register tokennuke, your agent can query indexed symbols and relationships locally instead of repeatedly opening whole files.
- Local code index queryable through 15 MCP tools
- Hybrid search and call-graph style navigation for 10 languages
- Faster agent turns on structural questions in your workspace
Recommended MCP Servers
Journey fit
Canonical shelf is build/agent-tooling because the primary value is equipping agents with a durable code index during implementation. agent-tooling fits MCP servers that extend what Claude/Cursor can query in the repo, not a single app feature.
How it compares
Local code-index MCP server, not a generic web-scraping or browser automation skill.
Common Questions / FAQ
Who is tokennuke for?
Solo and small-team developers who use MCP-enabled agents and want faster, structure-aware search across a multi-language codebase on their machine.
When should I use tokennuke?
Use it during active build and refactor work, or before review, when you need call graphs, definitions, and hybrid search without rereading entire trees each session.
How do I add tokennuke to my agent?
Install the PyPI package tokennuke, configure an stdio MCP server entry in Claude Code or Cursor pointing at that command, then restart the client so the 15 indexing tools appear.