An MCP snippet server for Claude Code
Point Claude Codeat The-Snip's hosted MCP server and it searches your team's reviewed base of snippets, API calls and docs mid-task, then files what it learns back — every write lands in review for a human to approve. Reads are free; every Free workspace gets 5 free agent writes, then Pro & Team ($8/user/mo) unlocks unlimited writes.
Give Claude Code a knowledge base it can read and write instead of re-deriving your team's patterns every session. The-Snip's hosted MCP server exposes your snippets, saved API calls and Markdown docs as nine tools — reads like search_base, get_item and list_collections, writes like create_snippet and update_item — over Streamable HTTP with a workspace API key. Claude Code searches your base mid-task and cites the canonical helper instead of generating a new variant; when it solves something novel, it files the result back. Every agent write lands in the review queue, so a human approves what becomes canon. Setup is one JSON block and takes about two minutes.
Looking for a shared, reviewed memory across sessions instead? A reviewed team memory for Claude Code
How it works
- Create a workspace and seed it with the snippets your team reaches for weekly — even ten good ones make the loop click on day one.
- Generate a workspace API key in Settings → API keys — keys are free on every plan, with 5 free agent writes (review-gated); unlimited agent writes need Pro & Team, $8/user/month.
- Add the server to your project's .mcp.json with the config below; Claude Code picks it up on the next session.
- Tell Claude Code when to use it — one line in CLAUDE.md: search the-snip before writing shared utilities; file new reusable patterns with create_snippet.
- Work normally. Claude Code calls search_base mid-task and reuses your canonical snippets, citing what it found.
- Triage the review queue: approve, edit or reject what the agent filed. Approved items become canon for every teammate — and every agent.
{
"mcpServers": {
"the-snip": {
"url": "https://the-snip.com/mcp",
"headers": { "Authorization": "Bearer snip_········_····" }
}
}
}Questions, answered.
What tools does Claude Code get?
Nine — reads: search_base, get_item, list_collections; writes: create_snippet, create_api, create_doc, create_webpage, update_item, create_collection. Reads return your workspace's approved canon; writes land in review, never directly in canon.
Which plan includes the MCP server?
Every plan can connect. Reads are free — any agent can search the base on the Free plan, and every Free workspace gets 5 free agent writes (review-gated). Unlimited agent writes need Pro & Team, $8/user/mo — or $80/user/year; there's no separate agent tier.
Do agents need their own seats?
No. Agents authenticate with workspace API keys, not seats — you pay per human. Create separate keys per agent or per developer so any one of them can be revoked without touching the rest.
What stops Claude Code from filling the base with junk?
The review gate. Everything an agent writes is marked in-review and stays out of canon until a reviewer approves it. Reject freely — a lean, trusted base beats a big, doubtful one.
Does this work with MCP clients other than Claude Code?
Yes — the server speaks standard Streamable HTTP MCP with Bearer auth, so Cursor, Windsurf, Claude Desktop or anything MCP-compatible can use the same config shape and the same nine tools.
Can Claude Code update existing items too?
Yes — update_item proposes a change to an existing snippet, request or doc, and the edit waits in review like any other write. Stale canon gets corrected through the same gate that admitted it.
Start your base — free.
Free: 25 items, no card — plus agent search over the REST API and the hosted MCP server, and 5 free agent writes (review-gated). Pro & Team: $8/user/mo — unlimited items, unlimited agent writes, export/import, team workspaces and review.