Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/valarpirai/code-graph/claude-mdgit clone --depth 1 https://github.com/valarpirai/code-graphWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00772 | $0.00772 |
| Opus 5 | $0.00386 | $0.00386 |
| Sonnet 5 | $0.00154 | $0.00154 |
| Haiku 4.5 | $0.00077 | $0.00077 |
Grade A, and why
code-graph CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 98 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Code Graph ingests a GitHub repo or ZIP, parses it with Tree-sitter, builds an OWL/RDF knowledge graph, and serves it via a FastAPI backend + React frontend. An MCP server exposes the graph as AI-callable tools.
Docs
- Architecture — system diagram, file trees, ontology, data layout, gotchas
- MCP Server — tools, transports, integration guide, security
Commands
Backend (Python 3.11+, uv)
Always use uv — never pip directly.
cd backend
uv sync # install / refresh deps
uv run uvicorn app.main:app --reload # dev server :8000
uv run pytest # full test suite
uv run pytest tests/test_parsing/test_java.py -v
uv run pytest -k "test_counts"
uv run pytest --tb=short -q
Frontend (Node 22 required — nvm use 22)
cd frontend
npm install
npm run dev # Vite dev server :5173
npm run build
npm run test # vitest (non-watch)
npm run test:watch
All services
./dev.sh start # backend :8000, frontend :5173, MCP server :8001
./dev.sh stop
./dev.sh status
./dev.sh logs
./dev.sh mcp # MCP server only
Docker
docker-compose up --build # backend :8000, frontend :80
Environment
# backend/.env
ANTHROPIC_API_KEY=sk-ant-... # required for wiki search + NL SPARQL
DATA_DIR=/data # default; override for local dev
AI features (wiki/search, sparql/natural) return HTTP 503 without ANTHROPIC_API_KEY.
API conventions
All REST routes follow: POST|GET /api/v1/projects/{project_id}/<resource>
Routers live in backend/app/api/. New routers must be mounted in backend/app/main.py.
Testing
Test layout:
backend/tests/
conftest.py autouse fixture: clears get_settings() LRU cache
test_config.py
test_analysis/ blast_radius, clustering, execution_flow, graph_to_networkx
test_api/ test_projects, test_graph, test_analysis_endpoints, test_wiki_endpoints
test_ingestion/ github, zip_handler, language_detector
test_models/ project
test_parsing/ one file per language (java, python, typescript, …)
test_rdf/ builder, graph_store, ontology
test_storage/ project_store
test_wiki/ generator, sparql_queries
test_ws/ indexing_ws
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 98 lines · 772 tokens per session scan A fb7bc8e3a6cd
code-graph CLAUDE.md is an instructions file published in the GitHub repository valarpirai/code-graph (1 stars, last pushed 4mo ago), licensed MIT. It adds 772 tokens to every session, about $0.0039 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
ontobricks CLAUDE.md
Claude Code instructions for databrickslabs/ontobricks, covering claude.md — operating contract for ontobricks, imported canonical rules, claude-only additions, skills (auto-invoked by description) and databricks-related work.
ontobricks AGENTS.md
AGENTS.md instructions for databrickslabs/ontobricks: Pointer file for AI coding agents (Codex, Gemini CLI, OpenAI Agents, etc.).
hugegraph AGENTS.md
Instructions for apache/hugegraph, covering agents.md, stack & modules, code search anchors, build and all modules.
tinkerpop AGENTS.md
Instructions for apache/tinkerpop, covering agents.md, primary guidance: agent skills, planning and memory: beads, canonical documentation and licensing and provenance.
code-graph AGENTS.md
Instructions for FalkorDB/code-graph, covering codegraph - agent instructions, architecture, data flow, directory structure and commands.
msagljs copilot-instructions.md
Instructions for microsoft/msagljs, covering copilot instructions for msagl-js, project overview, build, test, and lint, architecture and core data model (@msagl/core).