Creativity · MCP — client

MCP Fleak Client

Fleak is a serverless low-code platform for building AI workflows. Its nodes can act as MCP client stubs — calling out to MCP servers (vector DBs, SaaS tools, custom Python) as part of a workflow, or exposing Fleak flows as MCP tools back to an agent. It's especially useful for teams that want to chain MCP tools without writing glue code.

MCP facts

Kind: client
Ecosystem: anthropic-mcp
Language: Hosted (Python runtime under the hood)
Transports: http

Capabilities

Fleak flows can call MCP servers as steps
Published flows can be exposed back as MCP tools for agents
Built-in nodes for embeddings, vector DBs, and SQL

Install

Configured via the Fleak web console; no local install required.

Configuration

{
  "mcpServers": {
    "fleak-flow": {
      "transport": "http",
      "url": "https://api.fleak.ai/mcp/{workspace}",
      "headers": {
        "Authorization": "Bearer ${FLEAK_API_KEY}"
      }
    }
  }
}

Frequently asked questions

Is Fleak open-source?

Fleak is a commercial SaaS platform. It interoperates with open standards (MCP, JSON Schema) but the runtime is hosted.

Why use it over writing MCP servers directly?

For teams without Python/TypeScript expertise, Fleak's visual composer plus managed hosting is faster. You trade raw flexibility for speed-to-deploy.

Can it sit between Claude and internal APIs?

That's a common pattern: internal API -> Fleak flow (with auth, validation) -> exposed as an MCP tool -> Claude. Fleak handles the stitching and audit trail.

Sources

Fleak — low-code AI workflows — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20
MCP servers directory — accessed 2026-04-20