Skip to content
DADL

FAQ

Why is it called DADL?

Section titled “Why is it called DADL?”

DADL stands for Dunkel API Description Language — named after Axel Dunkel of Dunkel Cloud GmbH, who developed the format.

What problem does DADL solve?

Section titled “What problem does DADL solve?”

Today, a separate MCP server is built for every REST API — hundreds on GitHub, all following the same boilerplate pattern. None of them cover the full API; MCP servers are structurally always a subset. A .dadl file instead describes a REST API declaratively in about 30 lines of YAML — no code, no custom server, no deployment. LLMs can generate .dadl files in seconds.

What is DADL?

Section titled “What is DADL?”

DADL (Dunkel API Description Language) is a declarative YAML format that describes REST APIs for consumption by ToolMesh. Instead of building a dedicated MCP server for each API, you write a .dadl file and ToolMesh handles authentication, pagination, retries, and type safety automatically.

What is ToolMesh?

Section titled “What is ToolMesh?”

ToolMesh is a secure execution layer between AI agents and enterprise infrastructure. It reads DADL files and exposes APIs via Code Mode — two tools (list_tools + execute_code) that give any MCP-compatible AI agent access to entire APIs in roughly 1,000 tokens.

How does DADL differ from OpenAPI?

Section titled “How does DADL differ from OpenAPI?”

DADL is purpose-built for AI tool use, not for general API documentation. Key differences:

  • Minimal surface area — a DADL file describes only what an AI agent needs, not every possible API behavior
  • Built-in auth orchestration — credential injection is a first-class concept
  • Composites — multi-step workflows written in JavaScript, executed in a sandboxed runtime
  • Code Mode — instead of exposing hundreds of individual tools, ToolMesh generates TypeScript definitions and lets the agent write code against the API

Do I need to run my own MCP server?

Section titled “Do I need to run my own MCP server?”

No. ToolMesh is the MCP server. You configure it with one or more .dadl files and it exposes all defined tools and composites via the MCP protocol. No custom server code required.

Which AI agents work with ToolMesh?

Section titled “Which AI agents work with ToolMesh?”

Any agent that supports the Model Context Protocol (MCP) — including Claude Code, Claude Desktop, Cursor, Windsurf, and others. ToolMesh exposes a standard MCP endpoint that these clients connect to.

What authentication methods does DADL support?

Section titled “What authentication methods does DADL support?”

DADL supports several auth types out of the box:

  • Bearer tokens — static API keys or tokens
  • OAuth 2.0 — client credentials, authorization code, and other grant types
  • Header-based auth — custom headers like X-API-Key
  • Basic auth — username/password combinations

Credentials are stored in ToolMesh’s credential store and injected at request time — they never reach the AI agent.

Can I combine multiple APIs?

Section titled “Can I combine multiple APIs?”

Yes. ToolMesh can load multiple DADL files simultaneously, each describing a different API. Composites can even orchestrate calls across backends, letting you build workflows that span multiple APIs in a single tool invocation.

What is Code Mode?

Section titled “What is Code Mode?”

Code Mode is ToolMesh’s primary interaction model. Instead of exposing each API endpoint as an individual MCP tool, ToolMesh provides two tools:

  1. list_tools — returns TypeScript definitions for all available operations
  2. execute_code — runs agent-written JavaScript against the API in a sandboxed goja runtime

This keeps the tool count low while giving agents full API access.

Is DADL open source?

Section titled “Is DADL open source?”

The DADL specification is published under CC BY 4.0. The DADL registry at dadl.ai is open — anyone can contribute definitions. ToolMesh itself is a commercial product by Dunkel Cloud GmbH.

Where can I get help?

Section titled “Where can I get help?”