agenthold runs on your own machine — the client starts it, so there is no endpoint to ping. 271 installs a week from pypi. Last commit 7 Jul 2026.
Shared versioned state for multi-agent AI workflows
Today is the operative word: we check agenthold every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add agenthold -- uvx agenthold
{
"mcpServers": {
"agenthold": {
"args": [
"agenthold"
],
"command": "uvx"
}
}
}
[mcp_servers.agenthold]
command = "uvx"
args = ["agenthold"]
{
"mcpServers": {
"agenthold": {
"args": [
"agenthold"
],
"command": "uvx"
}
}
}
{
"mcpServers": {
"agenthold": {
"args": [
"agenthold"
],
"command": "uvx"
}
}
}
MCP State Sidecar: durable state persistence for multi-agent AI workflows
Persistent state for AI agent workflows. Save, resume, and replay multi-step workflows.
Durable agent-to-agent handoffs and shared scratchpad for multi-agent workflows.
BugBug workflows for AI agents workflows for AI agents
Inter-agent communication for multi-agent Claude Code workflows.
A standardized testing harness for MCP servers and agent workflows
Step-by-step observability for MCP agent workflows
Verification and memory layer for coding-agent workflows.
Answers built from our own checks of this server.