Tentra

@rdanieli

About Tentra

Tentra gives your AI coding agent memory. The code-graph indexer walks your repository with Tree-sitter locally, extracts symbols and call edges, and stores them in a persistent

#code-graph #memory #architecture #cursor #claude-code #persistent-context

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "tentra": {
      "type": "sse",
      "url": "https://trytentra.com/api/mcp?key=YOUR_API_KEY"
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is Tentra?

Tentra is a persistent memory layer for AI coding agents, built on two pillars: a code graph for indexing repositories and an architecture workspace for describing systems in natural language and generating interactive diagrams and code in 14 frameworks.

How to use Tentra?

Connect via SSE by adding a configuration entry to your MCP settings with the URL https://trytentra.com/api/mcp?key=YOUR_API_KEY (get your API key at trytentra.com/settings after GitHub sign-in). Alternatively, run npx -y tentra-mcp for a local stdio setup that authenticates via GitHub on first use.

Key features of Tentra

32 MCP tools across architecture, code graph, and enrichment categories.
Code graph indexes files, symbols, imports, and call edges once.
Architecture workspace produces interactive diagrams and production-ready code.
99.4% token reduction in benchmarks versus file re-reading.
Zero API key setup and zero LLM cost on Tentra's infrastructure.

Use cases of Tentra

Querying "Where is auth handled?" returns file paths and line ranges in one call.
Tracing function calls with "What calls this function?" via BFS on the call graph.
Investigating architectural decisions linked to ADRs with get_decisions_for.
Determining file ownership per team or person using get_ownership.
Finding code similar to a snippet via pgvector cosine similarity search.

FAQ from Tentra

How is Tentra different from re-reading source files?

Tentra reduces tokens needed for "where is X?" queries by 99.4% (156.8× ratio) by using a pre-built code graph instead of re-grepping source every session.

What are the runtime requirements?

Tentra can be used with zero install via SSE or with a local stdio setup using npx -y tentra-mcp. Authentication is done via GitHub.

Where does data live?

The README does not specify where data is stored, only that indices are built from your repository and accessed via Tentra's API.

What transports and authentication are available?

Tentra supports both SSE (with an API key) and stdio (with GitHub authentication). No API key setup is needed for the stdio path on your side.

Are there any known limits?

—

Frequently asked questions

How is Tentra different from re-reading source files?

Tentra reduces tokens needed for "where is X?" queries by 99.4% (156.8× ratio) by using a pre-built code graph instead of re-grepping source every session.