comcan-ctx 0.1.4

Context management for AI coding agents — state awareness + expertise memory.

🧠 ComCan — Context Manager for AI Coding Agents

ComCan gives AI coding agents (Cursor, Copilot, Claude Code) a persistent, enterprise-grade cognitive architecture.

Most developers solve the "AI context problem" by pasting random .md files or relying on the IDE's automatic vector search. These approaches fail in large enterprise repositories because the AI agent loses track of what branch you are on, what commits you just made, and what the strict architectural rules are.

ComCan solves this by installing an invisible cognitive architecture directly into your Git repository:

• 🔄 State Engine (The AI's RAM) — ComCan automatically installs Git hooks (post-commit, post-checkout) that silently run in the background. Every time you switch branches or make a commit, ComCan generates a highly optimized CURRENT_STATE.md file. It mathematically truncates your massive directory tree, diffs, and commit history to fit perfectly within the AI's token window without "Lost in the Middle" syndrome. When Cursor wakes up, it instantly knows your exact branch reality.
• 📚 Expertise Engine (The AI's Hard Drive) — Instead of writing massive, token-heavy .cursorrules files that cause merge conflicts across a team, ComCan uses lock-safe JSONL domain ledgers. When you execute comcan learn database "Always use soft deletes", it tracks this explicit rule. When an agent touches the database, it automatically queries this domain ledger.
• 🛡️ Enterprise Security — Built for corporate environments. Zero network access, zero HTTP calls, secrets are stripped from state files automatically, and no shell=True subprocess vulnerabilities exist. Standard IDE indexing (RAG) is great at finding where code is, but terrible at knowing why it's there or how branches differ.
• Indexing vs. State: IDE vector indexes don't understand that you just switched branches. ComCan's CURRENT_STATE.md is instantly updated via Git hooks to represent your exact branch reality.
• RAG vs. Rules: RAG discovers old code; it doesn't know what the new rules are. ComCan's expertise engine teaches agents the current architectural decisions.
• Static .md vs. Dynamic JSONL: Giant .cursorrules files cause merge conflicts and token bloat. ComCan uses domain-sharded, lock-safe JSONL ledgers that agents query surgically.

Philosophy: Why isn't domain learning "Automatic"?

Automatic codebase indexing (RAG) is prone to massive amounts of noise and hallucination. An AI cannot automatically deduce your team's architectural intent just by reading code. If a developer pastes a bad pattern from StackOverflow, an automated AI indexer treats that bad code as a "truth" to learn from.

ComCan treats AI knowledge like documentation. The comcan learn command acts as a conscious architectural ledger. When you or an agent establishes a rule ("Always use exponential backoff for the auth API"), it is explicitly recorded and saved to a Git-tracked .jsonl file. This guarantees that your AI agent is operating on 100% accurate, PR-reviewed instructions, free from automated scraping noise.

Quick Start

pip install comcan-ctx
cd your-project/
comcan init

That's it. ComCan will:

1. Create .comcan/ directory with a CURRENT_STATE.md context file
2. Install Git hooks to auto-update context on commits and branch switches
3. Create .cursorrules so Cursor reads the context automatically

Comprehensive Usage Guide

Step 1: Initialize

Run initialization at the root of your Git repository:

comcan init

This installs the Git hooks, configures .cursorrules, and builds your first Context State file.

Step 2: Create Logical Domains

Break your project down into logical domains (e.g., api, database, frontend, auth).

comcan add database
comcan add auth

Step 3: Record Expertise (The Core Loop)

Whenever you solve a tricky bug, establish a new convention, or finalize an architectural decision, record it immediately. You can do this yourself, or instruct Cursor/Claude to run this command for you:

# 1. Quick convention recording
comcan learn database "Always use WAL mode for SQLite to prevent locking"

# 2. Full record syntax (for detailed bug post-mortems)
comcan record api --type failure "Auth tokens not refreshed" --resolution "Added retry with exponential backoff"

Step 4: The AI Injects the Knowledge

You are now done! When you ask Cursor a question like "Write a new database fetch function", its custom .cursorrules file will silently instruct it to aggressively run:

comcan query database

The AI context window is instantly injected with all the recorded wisdom for that exact domain before it generates a single line of code.

Step 5: Code & Commit

As you write code, change files, and switch branches, ComCan's Git hooks will silently rebuild .comcan/CURRENT_STATE.md in the background. The AI will always know exactly what branch it is on and what the latest commits accomplished.

Step 6: Autonomous Agent Skills (Auto-Learn)

ComCan requires absolutely zero manual upkeep once initialized.

When you run comcan init, it natively generates instruction files for your AI agents:

1. .cursorrules and .cursor/rules/comcan.mdc (For Cursor users)
2. .agents/skills/comcan/SKILL.md (For Antigravity users)

These files explicitly instruct your AI to Autonomously run the comcan learn terminal command after it completes a complex coding task or bug fix. The AI will read your codebase, infer the architectural rules itself, and update the JSONL ledgers in the background entirely on its own!

How to trigger in Cursor:

Cursor natively discovers comcan.mdc in the .cursor/rules/ folder. You do not need to do anything. Simply ask Cursor to fix a bug in the Chat or Composer, and watch it organically launch the terminal and run comcan learn when it finishes writing the code.

How to trigger in Antigravity:

Antigravity natively discovers SKILL.md in the .agents/skills/comcan/ folder. When you are pair programming with Antigravity, it will read this skill folder at startup. It will autonomously execute run_command("comcan query") before it writes code, and run_command("comcan learn") after you approve its changes.

Monitor State

# Dashboard
comcan status

# Manual sync (usually automatic via hooks)
comcan sync

# Health check
comcan doctor

How It Works

1. comcan init        → Creates .comcan/, hooks, .cursorrules
2. You commit code    → Hook fires, CURRENT_STATE.md auto-updates
3. AI reads context   → Agent starts with full project awareness
4. AI solves problem  → You record the lesson with comcan learn
7. git push           → Teammates' agents get smarter too

Enterprise Features

ComCan is purpose-built to solve the "AI Cold Start Problem" for large engineering teams:

1. Iterative Knowledge Building 🧠

Instead of pasting the same rules over and over, developers use comcan learn to permanently record patterns, bugs, and architectural decisions into domain-specific ledgers. AI agents query these automatically.

2. Branch-Aware Context 🔀

CURRENT_STATE.md regenerates on every git checkout and git commit. If Dev A is on feature-auth and Dev B is on bugfix-ui, their agents see completely different, branch-accurate contextual states.

3. Conflict-Free Merging 🤝

ComCan configures .gitattributes to use merge=union for expertise JSONL files. Multiple developers can record new knowledge on different branches simultaneously without ever triggering a merge conflict.

4. Concurrent Agent Safety 🔒

Multiple agents running in parallel? No problem. The expertise engine uses advisory file-locking with atomic temp-file rotation. Multiple IDE tools or CI scripts can write to the same domain simultaneously without data corruption.

5. Token Budget Efficiency 📉

Dumping an enterprise codebase into an LLM window causes hallucinations and massive API costs. ComCan uses a multi-model token budget engine (o200k_base tokenizer aware) to mathematically allocate context window limits across the directory tree, commits, diffs, and expertise records.

Architecture

.comcan/
├── CURRENT_STATE.md           # Auto-generated (branch, commits, tree)
├── comcan.config.yaml         # Configuration
└── expertise/
    ├── database.jsonl          # Domain expertise (one per domain)
    ├── api.jsonl
    └── frontend.jsonl

Context Budget Profiles

ComCan uses only ~5% of the model's context window, leaving 90%+ for actual work, via the context_budget.py engine (tiktoken o200k_base).

Profile	Context Window	ComCan Budget	Target Models
standard	128k	~6,400 tokens	GPT-4o, Claude 3.5 Sonnet
large	200k	~10,000 tokens	Claude 4, Cursor default
max	1M+	~50,000 tokens	Gemini 2.5 Pro Max, Claude Opus

Note: Why use a budget? Without a token budget, dumping a large enterprise repo into an LLM causes severe "Lost in the Middle" syndrome and drains API credits. ComCan protects your context window by mathematically prioritizing recent commits, diffs, and surgical domain expertise.

Native AI Skills & PR Workflows

During comcan init, the CLI generates three native AI instruction files:

1. .cursorrules (Legacy IDE rules)
2. .cursor/rules/comcan.mdc (Cursor Rules format)
3. .comcan/comcan-skill.md (Portable generic AI skill)

Human-in-the-Loop Security: When an AI agent runs comcan learn to solve a problem, it writes directly to .comcan/expertise/domain.jsonl. Because this file is tracked in Git, the new "AI Skill" shows up in the Pull Request diff. If the AI hallucinates a bad rule, the Senior Engineer reviewing the PR rejects it. Bad AI knowledge never makes it to the main branch.

CLI Reference

Command	Description
comcan init	Interactive setup wizard
comcan sync	Regenerate context state
comcan add <domain>	Create expertise domain
comcan learn <domain> "lesson"	Quick-record a convention
comcan record <domain> --type <type> "content"	Full record syntax
comcan query [domain]	View domain expertise
comcan search <query>	Search all expertise
comcan prime [domains...]	Full context for agent injection
comcan status	Context dashboard
comcan forget <domain> <id>	Delete a record
comcan doctor	Health & security check

Security

ComCan is designed to never trigger security scanners:

• ✅ No shell=True — all subprocess calls target only git
• ✅ No setup.py — pure pyproject.toml, no post-install scripts
• ✅ No network access — zero HTTP calls, no telemetry
• ✅ No eval()/exec() — plain readable Python
• ✅ Secret scrubbing — API keys stripped before writing state files
• ✅ Path validation — all writes scoped to Git repo root

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT — see LICENSE.