ContextMesh Docs

Persistent semantic memory for AI agents. Store context once, retrieve it semantically from any agent, any session, any tool.

Base URL: https://contextmesh.dev

Quickstart

Get your API key at contextmesh.dev/#pricing, then store your first memory in 30 seconds:

python

pip install contextmesh

from contextmesh import Mesh

mesh = Mesh("cm_live_your_key")

# Store context
mesh.remember("prod DB is Postgres 15 on AWS us-east-1")

# Retrieve semantically
hits = mesh.query("what database are we running?")
for h in hits:
    print(h["score"], h["text"])

curl

# Store
curl -X POST https://contextmesh.dev/remember \
  -H "Authorization: Bearer cm_live_your_key" \
  -H "Content-Type: application/json" \
  -d '{"text": "prod DB is Postgres 15 on AWS us-east-1"}'

# Query
curl -X POST https://contextmesh.dev/query \
  -H "Authorization: Bearer cm_live_your_key" \
  -H "Content-Type: application/json" \
  -d '{"q": "what database are we running?", "top_k": 5}'

javascript

import { Mesh } from "contextmesh-mcp"

const mesh = new Mesh("cm_live_your_key")
await mesh.remember("prod DB is Postgres 15 on AWS us-east-1")
const hits = await mesh.query("what database are we running?")
console.log(hits)

~/.cursor/mcp.json or ~/.claude/mcp.json

{
  "contextmesh": {
    "command": "npx",
    "args": ["contextmesh-mcp"],
    "env": {
      "CM_KEY": "cm_live_your_key"
    }
  }
}

Restart Claude or Cursor after saving. The agent will automatically call remember and query as tools.

Authentication

All API requests require a Bearer token in the Authorization header.

http

Authorization: Bearer cm_live_your_key

API keys start with cm_live_. Get one at contextmesh.dev/#pricing. Keys are workspace-scoped — all data stored under a key belongs to that workspace only.

Keep your key private. Anyone with your key has full read/write access to your workspace.

Errors

Status	Meaning
401	Invalid or missing API key
404	Entry not found
429	Rate limit or monthly quota exceeded — upgrade your plan
400	Validation error — check request body
500	Server error — contact hello@contextmesh.dev

POST /remember

POST /remember Store a piece of context

Embeds the text and stores it in your workspace. Returns an ID you can use to delete this entry later.

Request body

Field	Type	Required	Description
text	string	required	The context to store. Max 10,000 chars.
tags	string[]	optional	Labels for filtering later. e.g. `["database", "infra"]`
source_agent	string	optional	Which agent wrote this. e.g. `"cursor"`, `"claude"`
confidence	float	optional	How confident you are (0.0–1.0). Default: 1.0

Response

{"id": "a1b2c3d4", "status": "stored", "chars": 48, "tags": ["database"]}

POST /query

POST /query Semantic search

Finds the most semantically relevant stored entries using vector similarity. Returns results ranked by score.

Request body

Field	Type	Required	Description
q	string	required	Natural language query. Max 2,000 chars.
top_k	int	optional	Max results to return (1–20). Default: 5
tag_filter	string	optional	Only return entries with this tag
min_score	float	optional	Minimum similarity score (0.0–1.0). Default: 0.3

Response

{ "query": "what database are we running?", "count": 2, "results": [ {"id": "a1b2c3d4", "text": "prod DB is Postgres 15...", "score": 0.921, "tags": ["database"], "source_agent": "cursor", "created_at": 1711234567}, {"id": "e5f6g7h8", "text": "DB backups run nightly...", "score": 0.784, "tags": ["database"], "source_agent": "claude", "created_at": 1711200000} ] }

DELETE /forget/{entry_id}

DELETE /forget/{entry_id} Remove an entry

Permanently removes a stored context entry by ID. Get the ID from /remember or /list.

Response

{"id": "a1b2c3d4", "status": "deleted"}

GET /list

GET /list Browse stored entries

Returns all stored entries in your workspace. Paginate with limit and offset.

Query parameters

Param	Type	Description
limit	int	Results per page (1–200). Default: 50
offset	int	Skip N entries. Default: 0
tag_filter	string	Filter by tag

Response

{"total": 42, "offset": 0, "limit": 50, "entries": [...]}

GET /usage

GET /usage Check plan usage

Returns this month's usage and your plan limits.

{ "workspace_id": "ws_a4f8c3d2", "plan": "team", "queries_used": 1842, "queries_limit": 1000000, "entries_stored": 234, "entries_limit": 5000000, "rate_limit_rpm": 500 }

Python SDK

bash

pip install contextmesh

Sync client

python

from contextmesh import Mesh

mesh = Mesh("cm_live_your_key")

# Store
id_ = mesh.remember("prod DB is Postgres 15", tags=["database"])

# Query
hits = mesh.query("what database?", top_k=5, min_score=0.3)

# Delete
mesh.forget(id_)

# List all
data = mesh.list(limit=50)

# Check usage
u = mesh.usage()

Async client

python

from contextmesh import AsyncMesh

mesh = AsyncMesh("cm_live_your_key")
await mesh.remember("prod DB is Postgres 15")
hits = await mesh.query("what database?")

JavaScript SDK

bash

npm install contextmesh-mcp

javascript

import { Mesh } from "contextmesh-mcp"

const mesh = new Mesh("cm_live_your_key")
await mesh.remember("prod DB is Postgres 15", { tags: ["database"] })
const hits = await mesh.query("what database?")

MCP — Claude & Cursor

Add ContextMesh as an MCP server so Claude and Cursor automatically remember and recall context during conversations.

~/.cursor/mcp.json or ~/.claude/mcp.json

{
  "contextmesh": {
    "command": "npx",
    "args": ["contextmesh-mcp"],
    "env": {
      "CM_KEY": "cm_live_your_key"
    }
  }
}

After saving, restart Claude or Cursor. The agent gets three tools: remember, query, forget.

No code required. The agent decides when to store and retrieve context automatically.

Plans & Limits

Free

$0/mo

1,000 queries/mo

10,000 entries

20 req/min

1 workspace

Solo

$19/mo

100k queries/mo

500k entries

100 req/min

1 workspace

Team

$99/mo

1M queries/mo

5M entries

500 req/min

10 workspaces

Enterprise

$599/mo

Unlimited queries

Unlimited entries

2,000 req/min

Unlimited workspaces

Upgrade or manage your plan at contextmesh.dev/billing.