🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman
CLI for generating AGENTS.md repo context
Unoplat Code Confluence scans a repository and produces the context files agents need to work from a source of truth. It is organized around per-repo knowledge, dependency tracing, and workflow commands for the CLI and related services.
Builders who run coding agents and want them grounded in the current shape of a repo.
You can keep each repo ready for agents with maintained context instead of rebuilding it from scratch every session.
What it does
AGENTS.md generation
Creates and maintains machine-readable `AGENTS.md` files per repository.
Repo context extraction
Captures entry points, interfaces, business logic, dependencies, and data models from the codebase.
Language-aware parsing
Uses modular Tree-sitter parsing to support Python and TypeScript codebases.
Agent skill
Includes a portable skill at `skills/unoplat-code-confluence-cli/SKILL.md` for compatible coding agents.
Workflow tooling
Provides Docker, Taskfile, Tilt, and hook files for running the system locally and in development.
How to get it
- 1Install it with the skills CLI
npx skills add https://github.com/unoplat/unoplat-code-confluence --skill unoplat-code-confluence-cli
- 2Run
ALPHA ████████████████░░░░ BETA
README
Unoplat-CodeConfluence
The Universal Code Context Engine
Always keep your agents ready with all the context required per repository.
Explore the docs »
Quick Start
·
Report Bug
·
Request Feature
Table of Contents
🎥 Demo: Agents.md-first Context
Like this? Star the repo ⭐
·
Watch on YouTube
·
Sample PR
What's in the demo: automatic AGENTS.md generation per repo and an org index that gives any coding agent a precise source of truth. See a sample PR created by the tool.
⚠️ The Problem
AI coding agents excel at greenfield projects (new codebases built from scratch) but struggle with brownfield codebases (mature, production systems with existing code).
Why? They burn most of their context window on exploration—searching files, tracing flows, connecting dots—leaving little capacity for actual implementation. By the time they're ready to code, they've hit the "dumb zone" where performance degrades sharply. And since they lack long-term memory, this cycle repeats with every conversation.
Multi-repo complexity makes it worse. When code is split across connected repositories, the agent exhausts its context just mapping dependencies between codebases—often before writing a single line.
Internal dependencies present another failure mode. The agent has no onboarding to proprietary systems, so it hallucinates usage patterns. Worse, when internal documentation has drifted from actual implementation, the agent trusts those "lies" and produces code that doesn't work.
The end result: slop code requiring heavy rework.
🎯 The Solution
Unoplat Code Confluence is the context engine for application development, organizing precise, up-to-date knowledge of your data models, entry points, endpoints, and more—so coding agents can deliver and maintain features 2–3x faster with higher quality.
AGENTS.md-first Context
Auto-generates machine-readable AGENTS.md files per repo to give coding agents a precise source of truth:
- Engineering Workflow — Canonical install/build/dev/test/lint/type_check commands plus key config files and their responsibilities
- Business Logic — Core application logic, domain entities, and database entities
- Entry Points & Interfaces — Main entry points, API endpoints, and external interfaces
- External Dependencies — Roles and responsibilities of external libraries
🌟 Core Principles
1. Precision First
- Extensible Language Support: Modular Tree-sitter based grammar extraction delivers consistent, accurate code context across all programming languages
- Extensible Framework-Aware Parsing: Specialized grammar engines recognize framework and library-specific patterns based on project dependencies
2. Context Engineering
- All important metadata about application—dependencies, inbound/outbound interfaces, domain models, and data store models—are identified and their relationships preserved
3. Enterprise-Grade Scalability, Reliability and Auditability
- Scalable, auditable and reliable processing powered by workflow orchestrator
🚀 Getting Started
Ready to enhance your development workflow?
Check out our Quick Start Guide.
🧠 Agent Skill
This repository includes a portable agent skill for the Unoplat Code Confluence CLI at skills/unoplat-code-confluence-cli/SKILL.md.
Install it with the skills CLI:
npx skills add https://github.com/unoplat/unoplat-code-confluence --skill unoplat-code-confluence-cli
The skill teaches compatible coding agents how to use ucc from PATH, help users install it with uv tool install "git+https://github.com/unoplat/unoplat-code-confluence.git#subdirectory=unoplat-code-confluence-cli" when needed, ingest repositories, and run the single-command ucc agent-md <repository_git_url> flow.
📊 Project Status
ALPHA ████████████████░░░░ BETA
🚀 Sprinting Toward Beta
Unoplat Code Confluence currently supports Python and TypeScript codebases.
We're heads-down closing the remaining blockers to reach Beta.
👉 Track our Beta blockers →
Every open urgent issue is actively being worked on. We're shipping fixes daily.
For detailed roadmap, language support status, and planned features, see our Product Roadmap.
Maintainers
|
Jay Ghiya
Contact: jayghiya@unoplat.io |
Book a call with me - Cal Link
💬 Product Feedback & Alpha Disclaimer
Unoplat Code Confluence is in alpha. We’re building for our own daily use first, prioritizing stability and bug fixes. We’re collecting feedback now and will act on it once the core is solid. Early adopters welcome. Expect rapid changes and rough edges.
- Discord: Join our community channel
- GitHub Issues: Create an issue for bug reports or feature requests
- GitHub Discussions: Start a discussion for broader conversations
Your feedback is invaluable as we work toward production readiness and helps us prioritize our roadmap to better serve the developer community.
License
Unoplat-CodeConfluence is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0) + COMMONS CLAUSE.
Files in the repo
- .githooks
- .github
- .well-known
- backlog
- images
- scripts
- skills
- unoplat-code-confluence-cli
- unoplat-code-confluence-commons
- unoplat-code-confluence-docs
- unoplat-code-confluence-frontend
- unoplat-code-confluence-ingestion
- unoplat-code-confluence-query-engine
- yak
- .gitignore
- AGENTS.md
- APP_CHANGELOG.md
- architecture.d2
- architecture.svg
- backlog_instructions.md
- CLAUDE.md
- Contribution_Guidelines.md
- Dockerfile.dev-workflow-python
- Dockerfile.dev-workflow-typescript
- GEMINI.md
- LICENSE
- local-docker-compose.yml
- prod-docker-compose.yml
- README.md
- setup-hooks.sh
- Taskfile.yml
- tilt-docker-compose.yml
- Tiltfile
- version.txt
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More tools
The best-benchmarked open-source AI memory system. And it's free.
Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io
Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 550+ contributors
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.