continuity-core
OfficialClick on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@continuity-corerecall my commitments from yesterday"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Pick up tomorrow where you left off today.
Continuity Core remembers a commitment, finds it in a later session, and gives your assistant the exact next action and finish line. Run it locally and connect any MCP client to http://127.0.0.1:43187/mcp.
Try it
Requires Node.js 24+ and pnpm.
pnpm install
pnpm startIn another terminal, use the included client built with the official MCP SDK:
pnpm smoke write my-dayThe response includes a commitment ID. Stop and restart the server, then open a new client process:
pnpm smoke read my-day 1Replace 1 with the ID returned by the write command. The client connects over Streamable HTTP, lists the tools, recalls the saved commitment, and calls resume_commitment for its next step.
Related MCP server: LedgerMem MCP Server
Use it from your MCP client
Point a Streamable HTTP connection at http://127.0.0.1:43187/mcp. The server exposes three tools:
remember_commitment: save a commitment with a caller chosen context, next action, and observable finish line.recall_commitments: find up to 20 commitments in that context, newest first. Call with{ "context": "my-day" }for the latest page. Ifnext_before_idis a number, pass it asbefore_idwith the same context to read the next page. Anullcursor means you have reached the oldest commitment.resume_commitment: retrieve one commitment as a next step card.
Set CONTINUITY_DB_PATH to choose the SQLite file. The default is .local/continuity.sqlite in the directory where the server starts. CONTINUITY_PORT changes the loopback port.
Run on Cloudflare Workers
The Worker serves the same three MCP tools at /mcp and keeps commitments in D1 across requests. Every /mcp request needs Authorization: Bearer <CORE_SHARED_SECRET>.
For local Worker development, put CORE_SHARED_SECRET in an untracked .dev.vars file, then run:
pnpm install --frozen-lockfile
pnpm exec wrangler d1 migrations apply DB --local
pnpm dev:workerConnect a Streamable HTTP client to http://127.0.0.1:8787/mcp with that bearer token. The included smoke client also accepts CORE_SHARED_SECRET from its environment and CONTINUITY_URL for the endpoint.
To publish under your Cloudflare account:
pnpm exec wrangler d1 create continuity-coreReplace the all-zero database_id in wrangler.jsonc with the ID returned by that command. Then apply the schema and set the secret before the final deployment:
pnpm exec wrangler d1 migrations apply DB --remote
pnpm exec wrangler secret put CORE_SHARED_SECRET
pnpm exec wrangler deploywrangler secret put prompts for the value and deploys a Worker version. wrangler.jsonc declares the secret as required, so wrangler deploy rejects a missing secret; the Worker also refuses /mcp when no secret is bound. The D1 migration creates the commitments table and an index for context pagination.
This setup uses the Workers Free limits of 10 ms CPU per request and 100,000 requests per day, plus the D1 Free allowance of 5 million rows read and 100,000 rows written per day with 5 GB total storage. The recall tool fetches at most 21 indexed rows per call. Free limits stop requests when exhausted; they do not upgrade the account.
Call Nemotron
Set NEMOTRON_API_KEY, then send a short prompt:
NEMOTRON_API_KEY=your-key pnpm nemotron "Give me one next step for my project."The development default uses DeepInfra's Nemotron 3 Super. Set NEMOTRON_BASE_URL and NEMOTRON_MODEL to use another OpenAI-compatible endpoint. The command prints the answer, token usage, and elapsed time.
License
MIT. See LICENSE.
This server cannot be deployed
Maintenance
Related MCP Connectors
Cross-tool persistent memory and context for AI assistants over MCP.
Persistent personal memory for AI assistants — save, search, and recall across every MCP client.
- mcpOAuthai.butlerbrain
Persistent memory for AI assistants. Save once; recall from Claude, ChatGPT, or any MCP client.
Hosted MCP memory and agent control plane for durable conversations, jobs, and operations.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to maintain context across conversation sessions by saving and retrieving summaries of key points from past interactions.334 npmMIT

LedgerMem MCP Serverofficial
AlicenseNot gradedqualityBmaintenanceEnables persistent memory storage and retrieval for MCP clients, allowing AI assistants to remember facts and context across conversations.16 npmMIT- AlicenseNot gradedqualityAmaintenanceDurable, user-keyed state management for stateless MCP servers. Enables agents to persist and resume state across conversations, clients, and devices.MIT
- AlicenseNot gradedqualityCmaintenanceEnables MCP apps to manage durable, model-defined goal phases with persistent state, checkpointed progress, and safe resumption after interruptions or lost turns, ensuring elapsed time never implies completed business work.1MIT