mon (মন)

Bengali for mind. A Cloudflare-native memory layer for AI tools.

One Worker. One D1 table. One Vectorize index. Any AI tool that speaks MCP can plug in.

Deploy in under 5 minutes. $0 to run.

---

Why mon exists

Most AI tools still treat memory as an add-on. mon makes memory a tiny, deployable primitive: ingest a thought, search it semantically, expose it over MCP, and plug it into any AI workflow.

What it does

• Capture a thought via POST /ingest
• Recall semantically via GET /search?q=
• Expose recall_memory and save_memory MCP tools for MCP-compatible clients and agent runtimes (via your Worker's /mcp endpoint).

No Supabase. No Postgres. Built natively on Cloudflare.

---

Stack

Layer	Technology
Compute	Cloudflare Workers
Structured storage	Cloudflare D1 (SQLite)
Vector search	Cloudflare Vectorize
Embeddings	Workers AI (@cf/baai/bge-small-en-v1.5)
Protocol	MCP (Model Context Protocol)

---

Quick start

1. Prerequisites

• Cloudflare account (free tier works)
• Wrangler CLI installed
• Node.js 18+

2. Clone and install

git clone https://github.com/ibraasif/mon.git
cd mon
npm install

3. Create D1 database

wrangler d1 create app-mon-db

Copy the database_id from the output and paste it into wrangler.toml.

4. Create Vectorize index

wrangler vectorize create mon-index --dimensions=384 --metric=cosine

5. Apply schema

wrangler d1 execute app-mon-db --file=schema.sql --remote

6. Set auth token

mon protects all endpoints with a bearer token. Set it as a Wrangler secret — never hardcode it:

wrangler secret put AUTH_TOKEN

You'll be prompted to enter the token value. Use a strong random string (e.g. a UUID or 32+ character random value).

7. Deploy

wrangler deploy

That's it. Your memory endpoint is live at your Worker URL.

---

API

All endpoints require Authorization: Bearer <your-token>.

POST /ingest

Store and embed a thought.

curl -X POST https://mon.your-domain.com/ingest \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <your-token>" \
  -d '{"text": "Your thought here."}'

Response

{ "id": "uuid", "status": "stored" }

---

GET /search?q=

Recall semantically similar thoughts.

curl "https://mon.your-domain.com/search?q=your+query" \
  -H "Authorization: Bearer <your-token>"

Response

{
  "results": [
    {
      "id": "uuid",
      "text": "Your thought here.",
      "created_at": "2026-04-24T03:27:38Z",
      "score": 0.75993013
    }
  ]
}

---

MCP

mon exposes two MCP tools over a single /mcp endpoint using the MCP Streamable HTTP transport (JSON-RPC 2.0). All MCP requests require the same bearer token.

Tool	Description
recall_memory	Semantic search over your stored thoughts
save_memory	Store a new thought or piece of information

Client compatibility

Client	Status	Notes
claude.ai (web)	✅ Working	Streamable HTTP + bearer token header
Claude Desktop	✅ Working	Via mcp-remote stdio bridge
Perplexity	❌ Not supported	Requires persistent SSE — incompatible with Cloudflare Workers

claude.ai

1. Go to Settings → Integrations → Add custom connector
2. URL: https://mon.your-domain.com/mcp
3. Transport: Streamable HTTP
4. Authentication: API Key → add header Authorization: Bearer <your-token>

Claude Desktop

Claude Desktop uses stdio, not HTTP. Use mcp-remote as a bridge.

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "mon": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mon.your-domain.com/mcp",
        "--header",
        "Authorization: Bearer <your-token>"
      ]
    }
  }
}

Config file locations:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Why not Perplexity?

Perplexity's Streamable HTTP client opens a persistent GET SSE stream after initialize for server-to-client notifications. Cloudflare Workers terminate long-lived connections — making this architecturally incompatible without Durable Objects. PRs welcome.

---

License

FSL-1.1-MIT — Source-available under FSL-1.1-MIT; each release converts to MIT after 2 years.

---

Roadmap

See ROADMAP.md for what's coming after v0.

---

Built by @ibraasif. Inspired by the problem OB1 solves — rebuilt from scratch for the Cloudflare ecosystem.