Repository · Quick start · Docs · Benchmarks · Discussions
Graphicle maps your code and remembers your project, for you and your coding agents.
It builds a graph of your repository (files, functions, classes, calls and imports)
and keeps your documents, decisions and past agent sessions word for word, so
answers come with their sources. One command, graphicle, runs it all, and
everything it stores stays on your machine.
brew install graphicle-ai/tap/graphicle
cd your-repo && graphicle setupSetup builds the code graph, connects the coding agents it finds (Claude Code,
Codex, Cursor, VS Code and Claude Desktop) and lists every file it changes.
Then ask your agent to "use graphicle to find where failed requests are
retried", or run graphicle ask in the terminal. It runs on macOS and Linux, with a
Windows preview;
docs/install.md
has the install script, the release archives and building from source.
|
One map of your code Search the graph, explain a symbol, trace the path between two symbols and see what depends on a change. The engine is a Rust port of Graphify. |
Memory with sources Recall returns cited excerpts without calling a model. Agents can propose memories that quote their sources, and a person approves each one. |
|
Built for agents MCP for Claude Code, Codex, Cursor, VS Code, Claude Desktop and any other MCP client, with session capture for Claude Code and Codex. |
Local first No account and no telemetry. Nothing goes to a model unless you set one up (every case). |
Also built in: one MCP server, graphicle mcp, with the code-graph tools
under Graphify's tool names (such as query_graph and shortest_path) next to
the memory tools; reviewed memories tied to the functions they are about (a memory goes stale when its function changes; edits elsewhere in the file usually leave it current); and recall across the repositories
of a workspace (graphicle context --workspace).
- Speed: on 13 codebases, a pre-release build of Graphicle 0.1.1's engine took 3.0 to 22.7 times less time than Graphify 0.9.91 in all 39 measurements (same machine, 3 CPUs each). The graphs differ slightly: node counts within 4.7%, edge counts within 2.0%.
- Memory: a later pre-release build used less peak memory than Graphify 0.9.91 in all 39 measurements, from 0.29 to 0.97 times its peak; its graphs differ from Graphify's too.
- Coding agents: with Claude Code and Opus 5.5, Graphicle ties plain Claude Code (grep) on accuracy in every test. On Graphify's own showcase corpus, the 0.1.1 agent defaults (a pre-release build) found 99.2% of the facts, against 94.7% for plain Claude Code (a tie) and 87.8% for Graphify 0.9.91, with fewer tokens than Graphify.
- Parity: the code-graph engine is checked by running Graphify's own tests against it and comparing the results with the pinned Graphify release it follows. The upstream test files that are not run yet are listed with the reason each is not run.
- Agent memory: seven memory systems and a build of Graphicle answered the same questions on three development sets, with one open model as every system's reader and judge (a proxy evaluator).
Every result, with its labels and caveats, is in docs/benchmarks.md, including the ties and the results that did not favour Graphicle.
Graphicle Core is free and open source under the Apache License 2.0. Contributions are welcome with a DCO sign-off (CONTRIBUTING.md). The code-graph engine is adapted from Graphify, credited in NOTICE. A commercial edition for teams is in private preview: start a discussion.
- Questions and ideas: Discussions
- Chat: Discord
- Bugs: open an issue
- Security problems: report them privately, never in a public issue
- Everything else: vikramlex@karlexai.com · @vikram_lex on X · karlex.ai
Graphicle is made by 
KarLex AI, Inc. Other product names are trademarks of their owners; no affiliation or endorsement is implied.
