Fiber AI MCP Server

What is Fiber AI MCP Server?

Fiber AI MCP Server (Core) is Fiber AI's hosted endpoint at https://mcp.fiber.ai/mcp that lets AI assistants discover and call any Fiber B2B data API operation, from company and people search to contact enrichment, through five meta-tools. It uses a Fiber API key.

Fiber AI describes itself as an agent-native B2B data platform for sales, recruiting and GTM operations. Teams use it to find companies and people, enrich profiles and reveal work contact details, through an API, MCP or a plugin for coding agents. Fiber runs several MCP servers; this Core server is the most general one, giving an agent access to the full catalog of 100+ API operations without loading a separate tool for each.

How to use Fiber AI MCP Server?

Generate an API key in the Fiber app (fiber.ai/app/api) and keep it in an environment variable such as FIBER_API_KEY, so the agent never sees the raw key. Fiber checks for the key in this order: an apiKey field in the request, an x-api-key header (recommended for IDEs and CLIs), then Authorization: Bearer.

Claude Code

From Fiber's docs:

claude mcp add --transport http fiber-ai-core https://mcp.fiber.ai/mcp \
  --header "x-api-key: $FIBER_API_KEY"

Run /mcp in a session to confirm the connection.

Claude Desktop and claude.ai

Open Settings (or Customize), then Connectors, then "Add custom connector", and paste https://mcp.fiber.ai/mcp. Fiber's docs say that when Claude shows authentication as None, you add an x-api-key request header. If you prefer a browser sign-in instead of a key, Fiber's separate V3 server (https://mcp.fiber.ai/mcp/v3) supports OAuth.

Cursor

In Settings → MCP → New MCP Server, choose HTTP, or add to .cursor/mcp.json:

{
  "mcpServers": {
    "fiber-ai-core": {
      "url": "https://mcp.fiber.ai/mcp",
      "headers": {
        "x-api-key": "<YOUR_FIBER_API_KEY>"
      }
    }
  }
}

VS Code

Fiber's docs reference the environment variable directly:

{
  "servers": {
    "fiber-ai-core": {
      "type": "http",
      "url": "https://mcp.fiber.ai/mcp",
      "headers": { "x-api-key": "${env:FIBER_API_KEY}" }
    }
  }
}

Other Fiber servers

• V2 (/mcp/v2): about ten curated tools for common flows; Fiber suggests most users start here.
• V3 (/mcp/v3): one tool per public operation, OAuth sign-in only.
• Lite (/mcp/lite): a smaller subset for search, counts and quick contact reveal.

Key features of Fiber AI MCP Server

• Official server from Fiber AI with documented setup for Claude, Claude Code, Cursor, VS Code, Codex, ChatGPT, Windsurf and Zed.
• Access to the whole Fiber API through one compact tool set, instead of hundreds of separate tools.
• Ranked endpoint discovery for a stated intent, with curated "tag packs" to narrow large catalogs.
• Full request and response schemas on demand before the agent makes a call.
• Company, people and job-posting data, profile snapshots and work email and phone reveal.
• Per-call cost reporting through a chargeInfo field in each response.

Fiber AI MCP Server tools

The Core server exposes five meta-tools that work as a discover-then-call loop:

• Find the right operation: search_endpoints returns the top recommended operations for a user intent and can be scoped by tags or a curated tag pack. list_tag_packs lists those packs (for example social media or enrichment), and list_all_endpoints lists every endpoint with optional tag or search filters.
• Inspect it: get_endpoint_details_full returns the complete schema for an operation ID, including required fields and example usage.
• Run it: call_operation executes any Fiber API operation by operationId with body, path and query parameters.

Use cases of Fiber AI MCP Server

• Outbound prospecting: "Find Series B fintech companies in New York with 50 to 200 employees."
• Candidate sourcing: "Search for senior backend engineers in Berlin who work at developer-tools companies."
• Contact enrichment: "Get the work email for this person's profile."
• Account snapshots: "Pull a live snapshot of this company and summarize its recent hiring."
• Building GTM automations: let a coding agent explore operations with search_endpoints and get_endpoint_details_full before writing integration code.

Troubleshooting Fiber AI MCP Server

• HTTP 402 errors: your organization is out of credits. Each MCP call charges the same credits as the matching API operation.
• HTTP 429 errors: MCP calls share your organization's per-minute rate limits with your scripts.
• Client tries to start OAuth: OAuth works only on V3. For Core, configure the x-api-key header before connecting so the client does not attempt an OAuth flow.
• Wrong or missing parameters: have the agent call get_endpoint_details_full before call_operation for operations with nested inputs.