📋 Available on: the same 2026 plans that include API access — Startup, Growth, Enterprise, Agency 2026, Agency Standard 2026, and Custom 2026. Not available on legacy or Free plans.
The Archive MCP server connects your AI assistant directly to your Archive data. Instead of writing API queries, you add Archive as a connector in your assistant once, then just ask in plain language — "find this creator", "pull the transcript for this video", "show my competitor brands" — and the assistant calls Archive for you.
MCP (Model Context Protocol) is an open standard supported by many AI assistants, including Claude (claude.ai, Claude Desktop, Claude Code), ChatGPT, Cursor, Windsurf, Cline, and VS Code.
What You Can Do With It
The MCP server can both read your workspace and make changes to it — just ask in plain language.
Read Your Data
Workspaces — list the workspaces you can access and see what each one tracks (hashtags, mentions, keywords, connected accounts).
Content — search tracked posts and videos, pull the underlying media, get video transcriptions, and resolve a post URL to its item in Archive.
Creators & social profiles — search creators and social profiles, open a single one, and pull a profile's historical engagement.
Competitors — list tracked competitor brands and pull their content.
Campaigns & more — read campaigns, saved views and filter presets, custom attribute definitions, and the status of long-running operations.
Make Changes
The MCP server can also act on your workspace, not just read it:
Collections — create, rename, or delete a collection (a hand-picked set of posts), and add or remove posts from it.
Saved Views — create, update, or delete a saved view (a saved set of filters) for content, creators, or social profiles, and add it to a group.
Groups — create, rename, or delete a group (a folder that organizes your saved views), reorder the views inside, and add views to it.
Content actions — import a post into Archive from its link, or queue a bulk engagement refresh on posts (this spends credits and requires the refresh-engagement feature flag to be enabled on your workspace).
🔵 New: whether a write action asks you to confirm first depends on the tool-permission settings each person sets in their AI client. Deleting a collection, saved view, or group only removes that item — never the underlying content, creators, or media.
How to Connect
In Archive, go to Settings → API → Connect an agent and pick your assistant — Claude (claude.ai), Claude Code, Claude Desktop, ChatGPT, or Other — then follow the steps shown for that assistant. The server URL is:
https://app.archive.com/api/v2/mcp
📋 The server URL isn't a page you open in your browser — it goes into your assistant's connector settings. Start in the assistant you want to connect (open Claude, ChatGPT, etc.), then add Archive there.
Browser sign-in (recommended)
One-click sign-in through Archive in your browser — no API token to copy or store, and access is tied to you personally. This is the flow for Claude (on claude.ai) and ChatGPT.
In your assistant:
Open Settings → Connectors.
Choose Add custom connector (or MCP server).
Paste the server URL above.
Click Connect and sign in with Archive. The connection stays live after that.
Config file
Claude Code and Claude Desktop also offer a config-file option, and it's how you connect any Other client (Cursor, VS Code, Windsurf, Cline, Zed, SDKs). A project or desktop .mcp.json bridges to the server with mcp-remote and your API token — handy for scripted or shared setups, or for pinning an agency token across an organization.
{
"mcpServers": {
"archive": {
"command": "npx",
"args": [
"-y", "mcp-remote",
"https://app.archive.com/api/v2/mcp",
"--header", "Authorization: Bearer ${ARCHIVE_TOKEN}"
],
"env": {
"ARCHIVE_TOKEN": "<your-api-token>"
}
}
}
}You can also copy the exact per-client snippet from Settings → API, or connect directly over SDK / HTTP using the server URL and an Authorization: Bearer <your-api-token> header. To generate a token, see Generating an API Token; full details at api-docs.archive.com/getting-started.
Managing Connections
Browser sign-in is per person — each user authorizes their own assistant under the workspace. You can see and revoke connections at any time under Authorized connections on the same Settings → API screen; anyone in the workspace can revoke any listed connection.
Reconnection Required for Write Permissions
If you connected the Archive MCP server before write actions were implemented, you'll need to disconnect and reconnect to grant write permissions. The reconnection updates your access scope from read-only to read+write.
Users who connect the MCP server after this update automatically receive full read+write permissions — no extra steps needed.
Good to Know
Reads and writes. The MCP server can read your workspace and make changes to it — deleting a collection, saved view, or group only removes that item, never the underlying content.
Multiple workspaces. If your access spans more than one workspace, tell the assistant which workspace to use (it can list them first). With a single-workspace token this is automatic.
Rate limit. Requests share the same limit as the API — up to 5 requests per second per workspace.
Agencies. An agency-level token works through the MCP server the same way it does on the API.
For the full technical reference and per-client setup snippets, see api-docs.archive.com/getting-started. If you have questions, chat with us or email [email protected]. We're happy to help.