Developers
The short answer
Everything needed to get CodeMesh running and to know exactly what it does with your code: setup in four steps, how syncing actually works, connecting any MCP-capable agent, the full data policy, and the benchmark data behind every number on this site.
Set it up
Quickstart
From nothing to an agent answering questions about your repository. Four steps, in the order they have to happen.
Install the extension
One extension for VS Code, Cursor and Devin, including offline builds for every platform and what to do when the marketplace is blocked.
The CLI, no editor needed
npm install -g @codemesh-app/cli, then login, sync and connect. Full command reference, where state lives, and the environment variables.
Connect your agent
What Connect writes, how the handshake proves it worked, and notes for Claude Code, Cursor, Windsurf, VS Code and any other MCP client.
Understand it
How sync works
The daemon, the re-parse and the graph update, from your save to your agent's answer, shown as it happens.
Data policy
What is read from your machine, where it lands, what is recorded about your use, how long it is kept, and what never happens.
CodeMesh vs the alternatives
Against RAG, vector search, repository packing and prompt optimization. What each one is actually good at.
The console
Repositories, the graph, the agent connection and the activity log. Every tool call is listed with a preview of its arguments, which is how you find out what your agent has been asking for rather than guessing.
An organisation admin also sees per-member usage, storage and cost, and which repositories were indexed. Never the contents of those repositories.
Where you land. Whether every repository synced clean, how large the graph has grown, how many seats are in use, and which agents currently hold read access.
Example workspace (Northgate Systems). The reduction shown is the published benchmark's 95.1%.
Guides and reference
AI coding cost guide
Where AI coding spend actually goes, and the practical levers for bringing it down.
Context optimization guide
A practical guide to giving AI coding agents precise context instead of entire files.
Glossary
Plain-language definitions for the AI coding and context-optimization terms you will run into.
FAQ
Straight answers about what CodeMesh is, how it works, and what its benchmark actually measured.
The numbers
Benchmark
All 98 questions from the published benchmark against a public repository, including the 11 it got wrong.
Methodology
How the benchmark was run, what counted as a token, and how measured cost was calculated.
Benchmark explorer
Filter and inspect every benchmark question, token count and verdict individually.
Cost calculator
Estimate what unnecessary repository context could be costing your team, on your own numbers.
Research
Research and analysis on AI coding costs, token consumption and context optimization.
Start with the quickstart
Install, sign in, connect, ask. The free plan covers the whole loop.