# Chive — how to connect Chive turns saved social content into a searchable library your AI can query. Generate a ck_ API key in Settings → AI Connectors, then paste it once into Claude or ChatGPT. Keys are long-lived. Shown once. Revoke anytime. MCP URL: https://chive-mcp-production.up.railway.app/mcp Add to Cursor (one-click, MCP URL only): https://cursor.com/en/install-mcp?name=chive&config=eyJ1cmwiOiJodHRwczovL2NoaXZlLW1jcC1wcm9kdWN0aW9uLnVwLnJhaWx3YXkuYXBwL21jcCJ9 After install, add a request header named Authorization with value Bearer YOUR_API_KEY (a ck_ key from Settings). The one-click link only pre-fills the MCP URL. MCP directories: see the section below. Auth is Bearer YOUR_API_KEY (ck_ key). No OAuth. OpenAPI (ChatGPT Actions): https://usechive.com/openapi.json Connect sheet (this file): https://usechive.com/mcp.txt Human connect: https://usechive.com → Settings → AI Connectors (sign in, generate an API key, paste it once) App: https://usechive.com ## Auth Authorization: Bearer YOUR_API_KEY Same ck_ key for Claude MCP, ChatGPT Actions, Cursor, and curl. No OAuth on the MCP URL today. ## Claude.ai 1. In Claude.ai, open Customize → Connectors → Add custom connector. 2. URL: https://chive-mcp-production.up.railway.app/mcp 3. Authentication: None. Add a request header named Authorization with value Bearer YOUR_API_KEY. 4. Turn Chive on in a chat. Tools: search_chive, ask_chive, get_item. If your workspace has no Request headers field, use Claude Desktop below. ## Claude Desktop 1. Claude Desktop → Settings → Developer → Edit Config. 2. Paste the JSON, then quit the app fully (Cmd-Q) and reopen. { "mcpServers": { "chive": { "command": "npx", "args": [ "-y", "mcp-remote@latest", "https://chive-mcp-production.up.railway.app/mcp", "--header", "Authorization:${AUTH_HEADER}" ], "env": { "AUTH_HEADER": "Bearer YOUR_API_KEY" } } } } ## Claude Code claude mcp add --transport http chive https://chive-mcp-production.up.railway.app/mcp Then set the Authorization header to Bearer YOUR_API_KEY in that server’s config. ## ChatGPT 1. In ChatGPT, create a custom GPT and open Actions. 2. Import https://usechive.com/openapi.json 3. Set authentication to API Key (Bearer) and paste the ck_ key only — ChatGPT adds the prefix. 4. Save. Ask: Search my Chive library for [topic]. Operations: search_library, get_item, ask_library. You can search and ask this user's Chive library. Use search_library to find saves, get_item for one save, and ask_library for a written answer. Always send compact true on search_library and compact=true on get_item. If a tool returns nothing, say so. ask_library is Pro and spends one Chive credit. ChatGPT custom GPT Actions is the path that works today with a ck_ key. ChatGPT MCP / app connectors typically want OAuth; Chive MCP is Bearer-only and does not add OAuth. ## Cursor Add to Cursor: https://cursor.com/en/install-mcp?name=chive&config=eyJ1cmwiOiJodHRwczovL2NoaXZlLW1jcC1wcm9kdWN0aW9uLnVwLnJhaWx3YXkuYXBwL21jcCJ9 1. Open Add to Cursor. Cursor prompts to add the Chive HTTP server with the MCP URL filled in. 2. After install, add a request header named Authorization with value Bearer YOUR_API_KEY (a ck_ key from Settings). The one-click link only pre-fills the MCP URL. 3. Or paste the JSON into Cursor Settings → MCP / mcp.json and reload the window. { "mcpServers": { "chive": { "url": "https://chive-mcp-production.up.railway.app/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } } ## MCP directories List Chive as a remote HTTP MCP server. MCP URL: https://chive-mcp-production.up.railway.app/mcp Auth: Authorization Bearer ck_ (or x-api-key with the same ck_ key). No OAuth. Official registry metadata: https://usechive.com/server.json Smithery / scanner card (Railway origin, no API key): https://chive-mcp-production.up.railway.app/.well-known/mcp/server-card.json GitHub listing files: server.json, smithery.yaml, and glama.json at the repo root. Smithery URL publish still prefers OAuth when auth is required; the Railway card is the Bearer bypass. This repo does not add OAuth. Do not put a live ck_ key in a directory listing. ## Codex codex mcp add chive --url https://chive-mcp-production.up.railway.app/mcp Add the Bearer header in Codex MCP settings. There is no Chive OAuth login. ## What you get after connect MCP tools: search_chive, ask_chive, get_item search_chive finds items in your library. ask_chive answers from what you saved. get_item loads one item by id. Same library over HTTP: Search POST https://usechive.com/api/library-search Get item GET https://usechive.com/api/items/{id} Ask POST https://usechive.com/api/chat curl https://usechive.com/api/library-search \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{"q":"recipes","compact":true}' Ask is Pro and spends one credit. ## First message Search my Chive library for [topic] and tell me what I already saved. ## Agent paste-prompt Read https://usechive.com/mcp.txt and finish connecting Chive. I will paste the API key. ## Not this X / Instagram / YouTube connect in Settings is how content gets into Chive. That is not the AI connector. Granova is https://granova.ai/mcp.txt Knock Knock is https://get-knockknock.com/mcp.txt