Sitrep
Real-time parallel development coordination for AI agents and humans.
README
Sitrep
Real-time parallel development coordination for AI agents and humans.
Sitrep solves the divergence problem: when multiple agents vibe-code on the same codebase in parallel, their architectural decisions diverge silently. Git's unit of sharing (commit/push) is too coarse for the rate at which agents make decisions. By the time one agent pushes, another has made 50 divergent choices. Sitrep makes this transparent in real time.
What It Does
Sitrep is a centralized coordination server that AI coding agents connect to via MCP (Model Context Protocol). It provides:
- Sessions -- who is working on what, right now, on which branch, with what intent
- Decisions -- architectural choices recorded and shared across all parallel streams (ORM, patterns, deps, API shapes)
- Claims -- which files are being actively modified, by whom, with rich context about what is changing
- Discussions -- agent-to-agent async conversations to resolve conflicting architectural decisions
- Conflict Detection -- scope overlap warnings, file contention alerts, merge risk scoring
- Escalations -- unresolved agent conflicts surfaced to humans via the dashboard
- Branch Topology -- divergence tracking and cross-branch file overlap analysis
Getting Started
Prerequisites
- Docker and Docker Compose
- Git
- An MCP-capable AI agent (Cursor, Claude Code, VS Code Copilot, etc.)
1. Start the Server
git clone <repo-url> && cd sitrep
docker compose up -d
This starts five services:
| Service | Default Port | Purpose | Env Var |
|---|---|---|---|
frontend |
13000 | React dashboard for human oversight | SITREP_FRONTEND_PORT |
server |
18000 | FastAPI + FastMCP server (agents + API) | SITREP_API_PORT |
worker |
-- | Celery beat + workers (reaper, analysis) | -- |
postgres |
15432 | All persistent state | SITREP_POSTGRES_PORT |
redis |
16379 | Heartbeats, TTLs, Celery broker, pub/sub | SITREP_REDIS_PORT |
Ports are offset from standard defaults to avoid conflicts with services already running on your machine. To customize, copy .env.example to .env and edit:
cp .env.example .env
# Edit .env to change ports, credentials, or connection strings
Verify everything is healthy:
curl http://localhost:18000/health
# {"status":"ok"}
curl http://localhost:13000/api/projects
# []
2. Configure Your Agent's MCP Connection
Add the sitrep MCP server to your agent's configuration.
Cursor -- add to .cursor/mcp.json:
{
"mcpServers": {
"sitrep": {
"url": "http://localhost:18000/mcp/mcp"
}
}
}
Claude Code -- add to your MCP config:
{
"mcpServers": {
"sitrep": {
"url": "http://localhost:18000/mcp/mcp"
}
}
}
Adjust the port if you changed SITREP_API_PORT in your .env.
3. Set Up .sitrep/config.yaml in Your Repository
Create a .sitrep/ directory in the target repo with a config.yaml that identifies the project:
mkdir -p .sitrep
cp /path/to/sitrep/templates/sitrep-config.yaml .sitrep/config.yaml
# Edit .sitrep/config.yaml to set your project_id and server_url
Agents read this file before their first MCP call to auto-configure project_id, repo_url, and other defaults. The MCP server also auto-resolves projects from repo_url if the agent passes it, so even without this file, agents that pass their git remote will get matched to the right project.
4. Install the Agent Protocol Rules
Copy the appropriate rule file into your repository so the agent knows how and when to use sitrep:
For Cursor:
mkdir -p .cursor/rules
cp templates/sitrep-protocol.cursor.mdc .cursor/rules/sitrep-protocol.mdc
For Claude Code:
# Append to CLAUDE.md or create it
cat templates/sitrep-protocol.claude.md >> CLAUDE.md
For VS Code Copilot:
mkdir -p .github
cp templates/sitrep-protocol.vscode.md .github/copilot-instructions.md
For any agent (generic):
cp templates/sitrep-protocol.agents.md AGENTS.md
5. Install Git Hooks (Optional)
Git hooks notify sitrep when commits and pushes happen, enabling divergence monitoring:
bash hooks/install.sh --server-url http://localhost:8000 --project-id my-project
6. Create a Project
Via the dashboard at http://localhost:3000 (Config tab), or via API:
curl -X POST http://localhost:3000/api/projects \
-H "Content-Type: application/json" \
-d '{"name": "my-project"}'
7. Seed Architectural Decisions (Recommended)
Before starting parallel work, record key decisions that all agents should follow. You can do this via the MCP tools or the dashboard. This is the highest-value setup step -- an hour of ADR seeding prevents days of merge pain.
Development
Running Tests
pip install -e ".[dev]"
pytest
Running Locally (Without Docker)
# Start PostgreSQL and Redis separately, then:
uvicorn sitrep.server:app --reload
# In another terminal:
celery -A sitrep.workers.celery_app worker --loglevel=info --beat
# Frontend:
cd frontend && npm install && npm run dev
Project Structure
src/sitrep/
server.py # FastMCP + FastAPI combined entry point
config.py # Settings (env vars / .env)
mcp/tools.py # 7 MCP tools (begin, decide, discuss, claim, query, status, end)
core/ # Business logic (sessions, decisions, discussions, claims, conflicts, topology)
storage/ # PostgreSQL (repositories.py) + Redis (redis.py) + SQLAlchemy (database.py)
workers/ # Celery tasks (stale session reaper, branch analyzer, escalation checker)
api/ # REST + WebSocket + git hook endpoints for the dashboard
models/ # Pydantic domain models (domain.py) + SQLAlchemy ORM (db.py)
frontend/ # React + TypeScript + Vite dashboard
templates/ # Agent protocol rule files for each IDE
hooks/ # Git hooks (post-commit, post-push) + installer
docs/ # Best practices and team culture guide
Agent Protocol Rule Files
The templates/ directory contains rule files for each supported agent platform. These are the most important files in the system -- they teach the agent when and why to call sitrep tools during its reasoning process.
| File | Platform | Where to install |
|---|---|---|
sitrep-protocol.cursor.mdc |
Cursor | .cursor/rules/sitrep-protocol.mdc |
sitrep-protocol.claude.md |
Claude Code | Append to CLAUDE.md |
sitrep-protocol.vscode.md |
VS Code Copilot | .github/copilot-instructions.md |
sitrep-protocol.agents.md |
Any agent | AGENTS.md at repo root |
Documentation
- Best Practices Guide -- team culture document covering branching discipline, commit semantics, decision-first thinking, claim etiquette, and session hygiene. Customize this for your team.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。