
Agent Coherence
@Cohexa
About Agent Coherence
Stop AI agents from silently overwriting each other's shared files — a single-host coherence guard, TLA+-verified.
Config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"stale-write-guard-fs": {
"command": "uvx",
"args": [
"--from",
"agent-coherence[mcp]",
"stale-write-guard-fs"
]
}
}
}Tools
No tools detected
Fetch the live tool list by running this server in a temporary sandbox using the button above.
Overview
What is agent-coherence?
A coherence guard for shared files: when two agents (or sessions) write the same artifact, the stale writer is denied instead of silently overwriting — then reacquires and retries. Same protocol across LangGraph, CrewAI, AutoGen, the OpenAI Agents SDK, plain files (CoherentVolume), and any MCP client (stale-write-guard-fs).
How to use it
MCP: pip install "agent-coherence[mcp]", then run stale-write-guard-fs. Library: pip install agent-coherence and use CoherentVolume, or the CCSStore LangGraph drop-in.
Key features
- MESI single-writer ownership + invalidation — a stale write is denied fail-closed
- Optimistic commit-CAS for concurrent same-key writers
- Read-generation fence for crash-reclaimed writers
- Five safety invariants machine-checked in TLA+ (
make tla-checkin CI) - MCP server (5 tools), LangGraph/CrewAI/AutoGen adapters, Claude Code plugin
Honest scope
Single-host today: concurrent same-key writers on one host are covered; cross-host fencing is on the roadmap, demand-gated.
Basic information
More Files & Storage MCP servers
PDF Reader MCP Server (@sylphlab/pdf-reader-mcp)
sylphlab📄 The PDF intelligence layer for AI agents — Agent Document Twin, evidence-first extraction, visual crops, OCR provenance, trust reports, and benchmark-gated releases. MCP server for Claude, Cursor, VS Code, and any MCP client.
PDF Extraction MCP Server (Claude Code Fork)
xraywuMCP server to extract contents from a PDF file
md-pdf-mcp (Markdown to PDF MCP Server)
seanivoreMarkdown To PDF Conversion MCP
Filesystem MCP Server (@sylphlab/filesystem-mcp)
sylphlab📁 Secure, efficient MCP filesystem server - token-saving batch operations with project root confinement

Google Drive
modelcontextprotocolModel Context Protocol Servers
Comments