Connect your AI agent to your project's living memory: the SoluCortex MCP is here

Your agent starts every session from scratch

If you build with Claude Code, Cursor or another AI agent, you know the hidden cost: every new session starts with no memory. The agent doesn't know why a decision was made, which convention the project follows, or what was learned yesterday. You know — and you re-explain it. Again.

The connector that gives it memory

The SoluCortex MCP connects your agent to your project's living memory using the Model Context Protocol, the open standard spoken by Claude Code, Cursor and a growing list of clients.

Four tools cover the full loop:

• solucortex_recall — at the start of a task, the agent retrieves the approved decisions, conventions, risks and architecture, ranked by semantic similarity and importance.
• solucortex_search — ad-hoc questions mid-task ("how is authentication implemented?").
• solucortex_remember — at close, the agent proposes the new memories. You moderate: typed, approved, traceable memories — governance, not a notes dump.
• solucortex_list_memories — quick inspection without semantic search.

Connect in 2 minutes — nothing to install

The remote server at mcp.solucortex.ai is live in production. In Claude Code, one command:

claude mcp add --transport http solucortex https://mcp.solucortex.ai/mcp \
  --header "Authorization: Bearer scx_your_api_key" \
  --header "X-Solucortex-Project: your-project-uuid"

Your API key travels with each request — the server stores nobody's credentials. Prefer running it locally over stdio? You can: the setup guide covers both options for Claude Code, Claude Desktop, Cursor and any MCP client.

We use it every day

SoluCortex is built using SoluCortex: every task in our own roadmap starts with recall and closes by proposing memories. That's how we resumed a 40-plus task deliverable across sessions without re-explaining context.

Start today

1. Create your account and your first project — you get the API key and Project ID.
2. Connect your agent with the command above.
3. Ask it to recall context for its first task.

Your agent stops starting from scratch. Connect your agent in 2 minutes →