All stories
Graphicle Ecosystem Community reports

Graphicle launches a Rust code graph with memory for coding agents

Version 0.1.0 combines code search, cited project context and reviewed memories in one MCP server for clients including Codex and Claude Code.

Sources (4)

Why now Graphicle published its first public release October 14 at 13:22 UTC, followed by a Show HN submission at 15:00 UTC. This is a dated release, rather than a repository activity signal.

A code graph and memory in the same install

Graphicle has released version 0.1.0, combining a Rust port of Graphify’s code-graph engine with searchable project documents, past agent sessions and reviewed memories. Its pitch is practical: give a coding agent relevant code and earlier decisions as excerpts with file-and-line citations, through one MCP server. The release documentation names Codex, Claude Code, Cursor, VS Code and Claude Desktop as supported clients.

The project says memories can be tied to code symbols and withheld when those symbols change. Agents can propose memories, but a person must approve them before recall uses them. That addresses a concrete problem with persistent context: an old decision can outlive the function it describes.

Local retrieval, with visible setup changes

The Apache-2.0 release provides macOS and Linux binaries; its Homebrew route is brew install graphicle-ai/tap/graphicle. Native Windows support is planned, and the README calls WSL2 untested. Code-only graph extraction and context retrieval need no model API key. Generated answers can use a configured hosted provider or Ollama; hosted-provider charges depend on that provider.

Running graphicle setup changes agent configuration and repository instructions, installs Git hooks and enables session capture. The documentation describes backups and graphicle disconnect to undo its additions. Core retrieval runs locally, while configured model calls and some optional graph operations can send material to a provider.

In the English Hacker News discussion, the developer acknowledges that the project’s memory benchmarks do not yet answer whether a coding agent completes more tasks with Graphicle. The release offers a concrete integration to evaluate, without establishing a general productivity gain.