Central Intelligence
Give your agent durable memory across coding sessions so decisions, facts, and shared team knowledge survive context resets.
Overview
Central Intelligence is a Build-phase MCP server for the agent-tooling subphase that adds persistent cross-session memory store-and-recall for AI agents via stdio npm transport.
What is this MCP server?
- stdio npm package central-intelligence-mcp (registry version 0.1.3)
- Store, recall, and share knowledge across agent sessions via Central Intelligence API
- Free API key via npx central-intelligence-cli signup
- Optional CI_API_URL for self-hosted deployments
- npm identifier: central-intelligence-mcp
- Required secret env: CI_API_KEY
Community signal: 1 GitHub stars.
What problem does it solve?
Every new chat forgets your stack choices, prior debugging outcomes, and product decisions, wasting tokens and repeating mistakes.
Who is it for?
Solo builders running long-horizon products with MCP agents who need external memory instead of brittle project markdown alone.
Skip if: Builders who want fully local-only memory with no API key or cloud account, or who only need ephemeral single-session context.
What do I get? / Deliverables
After configuring CI_API_KEY, your agent can write and retrieve durable memories that persist across sessions and optional team sharing.
- Working stdio MCP memory bridge with authenticated API access
- Ability for agents to persist and recall structured knowledge across chats
- Optional self-hosted routing via CI_API_URL when not using default cloud API
Recommended MCP Servers
Journey fit
Persistent agent memory is anchored where you wire tooling—integrations and MCP servers— even though recall helps every phase once installed. Agent-tooling is the canonical shelf for MCP servers that extend what assistants remember and retrieve between runs.
How it compares
Persistent memory MCP service, not a planning skill or vector DB you operate yourself.
Common Questions / FAQ
Who is Central Intelligence for?
It is for developers using MCP-enabled agents who need session-spanning memory and optional shared knowledge via the Central Intelligence API.
When should I use Central Intelligence?
Use it when you repeatedly re-explain project context, want agents to remember fixes and conventions, or collaborate with shared recall across sessions.
How do I add Central Intelligence to my agent?
Install the central-intelligence-mcp npm stdio server, run npx central-intelligence-cli signup for CI_API_KEY, set that env var in your MCP config, and add optional CI_API_URL only for self-hosted instances.