Goldhold
Persist decisions, preferences, and facts across agent sessions so your stack does not re-discover context every day.
Overview
GoldHold MCP is a MCP server for the Build phase that gives AI agents persistent memory to search, store, and recall across sessions.
What is this MCP server?
- Search, store, and recall persistent memory across AI agent sessions
- Stdio npm @goldhold/mcp-server (v1.5.1) plus remote streamable-http at mcp.goldhold.ai and relay.goldhold.ai
- Bearer GOLDHOLD_API_KEY from goldhold.ai/account for hosted remotes
- Designed for cross-session continuity rather than one-shot chat context
- Version 1.5.1
- Two streamable-http remotes: mcp.goldhold.ai and relay.goldhold.ai
- Npm package @goldhold/mcp-server with stdio transport
What problem does it solve?
Agents lose thread between sessions, forcing you to re-upload specs, preferences, and debugging notes.
Who is it for?
Indie builders running long-horizon agent workflows who need reliable recall across repos and sessions.
Skip if: One-off tasks with no continuity needs, or teams that already enforce memory via a single monolithic knowledge base with strict governance.
What do I get? / Deliverables
After registration, your agent can write durable memories and retrieve them later, shrinking repeated context setup.
- Searchable persistent memory store tied to your agent
- Configured stdio or remote MCP endpoint
- Reduced re-onboarding time at the start of each session
Recommended MCP Servers
Journey fit
Durable agent memory is foundational tooling you attach during build and reuse through ship, launch, and operate. Agent-tooling is the canonical shelf for MCP servers that extend recall, search, and store primitives for coding agents.
How it compares
Hosted persistent memory MCP layer, not a project planner skill or codebase indexer.
Common Questions / FAQ
Who is GoldHold MCP for?
Developers using Claude Code, Cursor, or similar agents who need memory that survives beyond a single chat session.
When should I use GoldHold MCP?
Use it whenever you start a new session but want prior decisions, credentials notes, or architecture facts available without manual paste.
How do I add GoldHold MCP to my agent?
Either install @goldhold/mcp-server with GOLDHOLD_API_KEY for stdio, or configure the streamable-http remote URLs with a Bearer API key from goldhold.ai/account.