Tentra — Claude Code MCP server: what it does, how to install
An MCP server that keeps a code graph for AI agents, maps architecture, and exposes call relationships before you change a symbol.
claude mcp add --transport sse tentra https://trytentra.com/api/mcpWhat Tentra — Memory for AI Coding Agents does
Tentra gives an agent a persistent model of a repository: files, symbols, imports, and call edges can be queried through MCP tools. It supports fuzzy symbol search, reference lookup, call-graph traversal, rename planning, hotspot reports, and snapshot comparisons. The same server can analyze a codebase into an architecture diagram, check diagram drift, and store contracts, ownership links, and architecture decisions.
Run npx tentra-mcp init --hook to generate IDE configuration, add a Git post-commit hook, and save the derived repository ID in .tentra/metadata.json. A generated configuration uses TENTRA_API_KEY; the local npx tentra-mcp mode can authenticate through GitHub. Codebase scanning requires the local server, while SSE is the remote connection option. That distinction matters when an agent must inspect files rather than query a hosted endpoint.
What it helps with
- index a repository and search its symbols
- trace every caller before changing a function
- prepare a rename plan for a symbol and its call sites
- compare an architecture diagram with the code
- rank highly connected code nodes
Example prompts for the agent
- Index this repository with Tentra and list the symbols with the highest fan-in and fan-out.
- Find every caller of
processPaymentand prepare a safe rename plan. - Check the architecture against the current code and report any drift.
How to install Tentra — Memory for AI Coding Agents
1Platform CLI: remote server
claude mcp add --transport sse tentra https://trytentra.com/api/mcp2Platform CLI
claude mcp add --env TENTRA_API_KEY=<TENTRA_API_KEY> --transport stdio tentra-memory-for-ai-coding-agents -- npx -y tentra-mcpReplace <…> with your own values
Platform CLI
codex mcp add tentra-memory-for-ai-coding-agents --env TENTRA_API_KEY=<TENTRA_API_KEY> -- npx -y tentra-mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"tentra-memory-for-ai-coding-agents": {
"url": "https://trytentra.com/api/mcp"
}
}
}{
"mcpServers": {
"tentra-memory-for-ai-coding-agents": {
"command": "npx",
"args": [
"-y",
"tentra-mcp"
],
"env": {
"TENTRA_API_KEY": "<TENTRA_API_KEY>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"tentra-memory-for-ai-coding-agents": {
"type": "sse",
"url": "https://trytentra.com/api/mcp"
}
}
}{
"servers": {
"tentra-memory-for-ai-coding-agents": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"tentra-mcp"
],
"env": {
"TENTRA_API_KEY": "<TENTRA_API_KEY>"
}
}
}
}Replace <…> with your own values
1Platform CLI: remote server
gemini mcp add -t sse tentra-memory-for-ai-coding-agents https://trytentra.com/api/mcp2Platform CLI
gemini mcp add -e TENTRA_API_KEY=<TENTRA_API_KEY> tentra-memory-for-ai-coding-agents npx -y tentra-mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Tentra — Memory for AI Coding Agents
How do I install Tentra — Memory for AI Coding Agents in Claude Code?
Run claude mcp add --transport sse tentra https://trytentra.com/api/mcp, then reload Claude Code.
What does Tentra — Memory for AI Coding Agents do?
It maintains a queryable code graph for an AI agent and provides tools for architecture analysis, reference tracing, drift checks, and refactor planning.
What is required for codebase scanning?
Use the local server with npx tentra-mcp. SSE is the remote option, while a generated configuration uses TENTRA_API_KEY.
