DOCUMENTATION / v0.16.1
A clearer project.
A ready agent.
hnindex gives your AI agent project context and code locations backed by source evidence. Start locally; add semantic retrieval when you need it.
workspace_contextlocate_codegraph_contextStart with what you need
Know the workspace
Project purpose, live Git state and a declared task stored in SQLite.
Explore guideFind the right location
Files, symbols, line evidence and index freshness before editing.
Explore guideFollow the code graph
Multilingual AST relationships, callers and references.
Explore guideChoose your embeddings
Ollama, OpenAI, Voyage, Gemini and compatible APIs.
Explore guideYour first session
Run this from your project directory. Node 22+ is enough for the offline multilingual graph workflow.
npx -y hnindex-cli init --mcp claude
// Ask your MCP agent to run:
index_code_graph(path: "/your/project", project_name: "my-app")
workspace_context(project_name: "my-app", task: "Add session expiry")
locate_code(project_name: "my-app", symbol: "verifySession")Workspace context, tasks and the code graph live in SQLite. Semantic/hybrid search needs Qdrant and an embedding provider. Cloud providers receive the code chunks you choose to embed.
Context with evidence
Tasks are declared by the agent, rather than inferred from Git. The code graph now supports multiple AST language profiles; dynamic calls and external dependencies can remain unresolved. Check freshness before editing and use filesystem search for missing, excluded or stale code.