Architecture

Internal architecture for contributors and extenders

Camel-Kit is built on a 4-layer architecture designed for AI agent composability, cross-agent portability, and token efficiency. The system is designed around the principle: “The prompt is the product” — Camel-Kit ships instructions, not implementations.

Four Layers

Layer 1   AGENTS.md Routing
Cross-agent equalization
Layer 2   Skills (Markdown Instructions)
Composable, reusable guides
Layer 3   MCP Servers
Real-time verification & search
Layer 4   Graph CLI
Code intelligence (optional)

Graph CLI has 14 subcommands:

  • Analysis: stats, find, neighbors, path, subgraph
  • Camel-specific: route-flow, impact, route-topology, dead-code
  • Normalization: project-norms, project-context, route-context
  • Output: generate, visualize

Learn more about graph intelligence →

The Prompt Is the Product

Camel-Kit’s architecture embodies a key principle: the prompt is the product. Unlike traditional code generators that ship Java/Python implementations, Camel-Kit ships:

  • Markdown guides that instruct AI agents how to generate code
  • MCP tool definitions for real-time verification
  • Graph parsers for code analysis (optional)

This means:

  • No vendor lock-in — Skills work on any agent (Claude, Gemini, Qwen, etc.)
  • Easy customization — Edit Markdown files to change behavior
  • Version-independent — No recompilation when Camel versions change
  • Transparent — Users can read the exact instructions agents follow

Progressive Disclosure

Skills use progressive disclosure to minimize token usage:

  1. Metadata (always loaded) — Skill name, description, trigger patterns (~50 tokens/skill)
  2. SKILL.md (on trigger) — Main skill logic, loaded only when invoked (~500-2000 tokens)
  3. Guides (as needed) — Shared utilities, loaded only when referenced (~100-500 tokens each)

Example flow:

User: "Create a Camel project for order processing"
  → Agent sees "start a new Camel project" in AGENTS.md
  → Agent loads /camel-project/SKILL.md
  → SKILL.md references {{GUIDE:shared/camel-constitution.md}}
  → Agent loads constitution guide
  → Agent generates project requirements

This progressive loading keeps context usage minimal while maintaining full catalog coverage.

Context Efficiency

By combining progressive disclosure and MCP on-demand queries, Camel-Kit avoids loading full component catalogs into the agent’s context. This enables:

  • Faster agent responses — less context to process per turn
  • Support for smaller models — fits within constrained context windows
  • Cost reduction — fewer input tokens per request
  • Full coverage — every component is still verifiable via MCP

Next Steps