code_graph_agentic_observable_coding
llms.txt1 starsChanged 32 days agoApache-2.0View on GitHub
A C4 code graph tool plus a ready-made agent layer (agents, skills and playbooks for Claude Code) that lets teams of AI coding agents plan, build and review software against a shared, live ledger. Contracts between classes, tasks, ADRs and EVAL verdicts are recorded in the graph and enforced by the tool, and people watch every change in a web UI. Python 3.10+, SQLite, FastAPI, Cytoscape.js. Apache License 2.0. Source: https://github.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding Author: aishwaryajayanth1820-git. Credit: "Based on Code Graph by aishwaryajayanth1820-git".
In this repository4 other files it gives its agents.
- AGENTS.mdAGENTS.md
- c4-designSkill
- code-graphSkill
- orchestrate-cycleSkill
The file
Files
# Code Graph: observable, governed agentic coding > A C4 code graph tool plus a ready-made agent layer (agents, skills and playbooks for Claude Code) that lets teams of AI coding agents plan, build and review software against a shared, live ledger. Contracts between classes, tasks, ADRs and EVAL verdicts are recorded in the graph and enforced by the tool, and people watch every change in a web UI. Python 3.10+, SQLite, FastAPI, Cytoscape.js. Apache License 2.0. Key ideas: - The design is decomposed into C4 nodes (components, service classes, classes, interfaces, config and utility classes) before code exists. Every call between classes is a CALLS edge with an exact signature and purpose. - Governance policies (DESIGN_LOCK, LOCKED_NODE, OWNERSHIP, CONTRACT_CHANGE, STATUS, APPROVER) refuse unapproved changes with exit code 3 or HTTP 409. The change then goes through an architecture decision record (ADR). - Reviewer agents (architecture, application design, code, security, performance) record a verdict per node and EVAL. First-pass rates show which rules coders follow, and the least-followed EVALs become lessons in the coder playbook. - Deterministic, ArchUnit-style checks and platform-maturity metrics (instability, abstractness, distance from the main sequence) run over the graph. - It works beside any agent harness through the `cg` CLI (JSON output) or the HTTP API. Source: https://github.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding Author: aishwaryajayanth1820-git. Credit: "Based on Code Graph by aishwaryajayanth1820-git". AI use: AI crawlers, search engines and agents may crawl, index and summarize this site and the repository, and may use the author's content to train or improve AI models. Bundled third-party libraries keep their own licenses. ## Docs - [README](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/README.md): overview, quick start, governance model, using it with other agent harnesses - [Tool README](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/code_graph_tool/README.md): module map and web UI - [CLI reference](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/skills/code-graph/reference/cli.md): every `cg` command - [Data model and governance](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/skills/code-graph/reference/data-model.md): node fields, edge kinds, policies, checks, metrics - [Seed format](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/skills/code-graph/reference/seed-format.md): the contract-first C4 seed JSON - [AGENTS.md](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/AGENTS.md): instructions for coding agents working on the repository ## Agent layer - [code-graph skill](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/skills/code-graph/SKILL.md): how any agent reads and records the graph - [c4-design skill](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/skills/c4-design/SKILL.md): turning a plan into a contract-first C4 design and EVALs - [orchestrate-cycle skill](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/skills/orchestrate-cycle/SKILL.md): the code, review and refine cycle with parallel coders - [Coder agent](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/agents/coder.md): implements one task against the recorded contracts - [Baseline EVALs](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/.claude/skills/c4-design/reference/baseline-evals.json): the starting EVAL catalog ## Optional - [Sample seed](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/code_graph_tool/examples/sample_seed.json): a notification platform used by the tests and the demo - [Demo script](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/code_graph_tool/examples/demo.py): replays an orchestrated build, including blocked changes and ADRs - [License](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/LICENSE) and [NOTICE](https://raw.githubusercontent.com/aishwaryajayanth1820-git/code_graph_agentic_observable_coding/main/NOTICE)
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

