Skip to content
You must be logged in to sponsor anvanster

Become a sponsor to anvanster

@anvanster

anvanster

I build CodeGraph (https://github.com/codegraph-ai/CodeGraph) — a semantic code graph that gives AI agents structured understanding of codebases instead of blind grep.
Accompanied by Tempera (https://github.com/anvanster/tempera.git) - MCP Server for Self-Evolving Agents via Reinforcement Learning on Episodic Memory

It parses 38 languages via tree-sitter, builds a persistent graph of functions, calls, imports, and dependencies, and exposes it through 41 MCP tools that work with
Claude, Cursor, Windsurf, and any MCP-compatible client.

What CodeGraph does for developers:

  • "Who calls this function?" → one tool call, not 15 greps
  • "What breaks if I rename this?" → blast radius in seconds
  • "Generate architecture docs" → from live code, always fresh
  • "Remember this debugging insight" → persists across sessions

What Tempera does for developers:

  • "Have I solved this before?" → pulls past episodes with the same error fingerprint, ranked by what actually worked
  • "What's the playbook for this kind of task?" → reasoning template extracted from successful (task_type, domain) clusters, before you start
  • "Where do I tend to be wrong in this codebase?" → top correction categories anchored to files you're about to touch
  • "What did I figure out yesterday?" → daily reflections + cross-day patterns, auto-authored during the dream cycle
  • "Is this claim still true?" → scope-aware decay (forever vs. crate-version vs. workaround) keeps stale advice from drowning out fresh signal
  • "Which of these two memories should I trust?" → contradiction probe with Wilson CI flags the conflict instead of letting both win retrieval
  • "How calibrated am I really?" → per-(task_type, project) bucket showing how often declared "success" survives to verified-stable

Where we are today:

  • 800+ VS Code installs, 370+ weekly npm downloads
  • Actively maintained — shipping features weekly
  • Cross-platform: macOS, Linux, Windows
  • Free and open-source (Apache-2.0)

What sponsorship supports:

  • Build infrastructure (3 cross-platform build hosts)
  • Continued full-time development as a solo maintainer
  • Keeping the community edition free with no feature gates

Every dollar goes directly into making AI agents better at understanding code. No VC, no committee — just one developer shipping as fast as possible.

@anvanster

"Cover the build hosts" — CodeGraph ships on 4 platforms from 3 build machines (old Linux, Windows, macOS). Right now they run on my personal infrastructure. 20 supporters at $5/mo covers the hosting costs and means the cross-platform builds aren't subsidized by my coffee budget anymore. First milestone toward sustainability — after this, every new sponsor funds features instead of infrastructure.

Featured work

  1. codegraph-ai/CodeGraph

    CodeGraph builds a semantic graph of your codebase — functions, classes, imports, call chains — and exposes it through 42 MCP tools, 38 languages, a VS Code extension, and a persistent memory layer…

    C 25
  2. anvanster/tempera

    MemRL: MCP Server for Self-Evolving Agents via Reinforcement Learning on Episodic Memory

    Rust 4

Learn more about sponsoring developers and organizations.

Report abuse

0% towards 20 monthly sponsors goal

Be the first to sponsor this goal!

Select a tier

$ a month

Choose a custom amount.

$5 a month

Select

Get a Sponsor badge on your profile

$12 a month

Select

Your name will be added to the "Supporter" section of one of my personal projects of your choice. Please contact me by email where you would want this to show up.

$100 a month

Select

Your company logo will be added to the "Supporter" section of one of my personal projects on GitHub and my website and I will prioritize your issues, within reason. You chose the projects.