Meridian
Add persistent session memory, coordinated tasks, and a human-in-the-loop queue so long-running Claude Code or Cursor sessions do not lose context between turns.
Overview
Meridian MCP is a MCP server for the Build phase that provides persistent memory, task coordination, and a HITL queue tailored to AI coding sessions.
What is this MCP server?
- Persistent memory across AI coding sessions
- Task coordination primitives for multi-step agent work
- Human-in-the-loop (HITL) queue for approvals and blocked steps
- Hosted SSE remote at https://usemeridian.us/mcp/sse plus npm stdio package @meridianmcp/mcp (v1.0.2)
- GitHub source: github.com/meridianmcp/Meridian
- Published server version 1.0.2
- npm package identifier @meridianmcp/mcp with stdio transport
- SSE remote URL https://usemeridian.us/mcp/sse
What problem does it solve?
Long agent coding runs lose thread context, duplicate work, and skip human approval gates because nothing persists tasks or memory outside the chat window.
Who is it for?
Solo builders running multi-hour Claude Code or Cursor sessions on one codebase who need durable memory and explicit approval checkpoints.
Skip if: One-shot snippet generation with no ongoing project, or teams that already enforce memory via a full PM stack and do not want another coordination layer.
What do I get? / Deliverables
Sessions retain coordinated tasks and retrievable memory, and blocked steps land in a HITL queue you can clear before the agent continues.
- Cross-session persistent memory usable by the agent on your project
- Coordinated task state for multi-step implementation runs
- HITL queue entries for approvals and blocked operations
Recommended MCP Servers
Journey fit
Meridian is agent infrastructure used while you build—memory and task coordination are core Build / agent-tooling concerns for solo devs shipping with AI pair programmers. Agent-tooling is the shelf for MCP servers that extend the coding agent runtime itself (memory, queues, coordination), not app feature code.
How it compares
Agent session memory and HITL MCP—not a Git host, issue tracker, or deployment MCP.
Common Questions / FAQ
Who is Meridian MCP for?
Indie developers and agent-first builders who want persistent memory, task tracking, and human approval queues inside AI coding workflows.
When should I use Meridian MCP?
Use it during Build when sessions span many files and days, you need remembered decisions, or you want HITL gates before risky edits or commands run.
How do I add Meridian MCP to my agent?
Add either the SSE remote https://usemeridian.us/mcp/sse or install npm package @meridianmcp/mcp with stdio transport in your MCP client configuration.