Camel Kit
Camel Kit adds structured workflows for Apache Camel integration development to AI coding assistants. It guides an assistant through the integration lifecycle: requirements and design, planning, implementation, testing, and runtime verification, and it checks every component and option against the Camel catalog while it does so. It is available as a standalone command-line tool and as a Camel JBang plugin, and it generates the same skills for several coding agents.
This documentation follows the development branch of Camel Kit (0.4.2-SNAPSHOT). The latest release is Camel Kit 0.4.1 with Knowledge MCP 0.0.2, which target Camel 4.22.1; features that are not in the release are marked as development snapshot only.
What it does
-
AI-guided design. Slash commands run a structured interview and produce a versionable design specification.
-
Platform migration. Detects the source platform and converts MuleSoft 3.x and 4.x, Microsoft BizTalk, Camel 2.x and 3.x, and JBoss Fuse projects to Camel 4 with YAML DSL routes.
-
Catalog verification over MCP. Components, endpoint options and infrastructure
forage.*keys are verified against the Camel catalog and the Forage catalog at design and review time. -
Graph intelligence. Nine parsers and two post-processors build a dependency-injection-aware property graph of the codebase for analysis and migration context.
-
Runtime verification. Build, startup and test checks run in the real environment; failures are classified, fixed within a bounded retry budget, and reported.
-
Several AI agents. The same skills are generated for eight coding agents (see Supported AI agents).
The workflow
Entry: /camel-start routes to migration, planning,
execution, validation, debugging or
greenfield work based on context
Greenfield: /camel-brainstorm then the shared stages below
Migration: /camel-migrate then the shared stages below
Shared: /camel-plan -> /camel-execute -> /camel-validate
|- implements (camel-implement)
|- verifies (camel-verify, internal)
'- tests (camel-test)
Utilities: /camel-ship Technology Preview: delegates to the
local Ship controller
/camel-knowledge documentation Q&A
/camel-debug standalone troubleshooting | Command | Purpose |
|---|---|
| Entry point. Routes migration sources to migrate, approved designs to plan, ready plans to execute, generated routes to validate, broken projects outside a pipeline to debug, and other work to brainstorm. |
| Interactive design session that produces business requirements and a design specification. |
| Migration from MuleSoft, Microsoft BizTalk, legacy Camel or JBoss Fuse to Camel 4. |
| Reviews the approved design and creates an implementation plan with wave analysis for parallel execution. |
| Orchestrated execution: probes the environment, implements tasks in dependency waves, runs adversarial, spec and quality reviews, and performs the applicable smoke and runtime verification. |
| Final static quality gate, or a standalone project check: schema, endpoints, quality, security, anti-patterns and constitution compliance, without modifying routes. |
| Standalone troubleshooting outside a pipeline for broken routes, build failures, startup errors and runtime exceptions. |
| Technology Preview. Thin entry point for the local Ship controller, which owns the end-to-end run with configurable oversight ( |
| Documentation Q&A: semantic search over the Camel documentation, CVE advisories, release notes and the component catalog. |
See Stages, approval gate and Iron Laws for the stages, the approval gate and the Iron Laws, and Command reference for every option.
Supported AI agents
camel-kit init --ai <target> generates the project layout for one agent. Each row is an alternative layout, not one combined tree. All targets share the same Markdown skills, adapted through agent-specific traits and native configuration.
| Agent | Init flag | Primary generated assets | MCP configuration |
|---|---|---|---|
Anthropic Claude Code |
|
|
|
IBM Bob 2 (default) |
|
|
|
Google Antigravity |
|
|
|
OpenAI Codex CLI |
|
|
|
GitHub Copilot CLI |
|
|
|
Pi |
|
|
|
Qwen |
|
|
|
OpenCode |
|
|
|
See Supported AI agents for per-agent setup notes.
How it is built
Camel Kit has four layers:
-
Agent instructions. A generated
AGENTS.md(or the agent’s equivalent) routes requests to the right skill; this is the cross-agent equalization layer. -
Skills. Composable Markdown instructions, loaded progressively: metadata first, the skill when triggered, shared guides only when referenced.
-
MCP servers. Catalog verification, Citrus test guidance and documentation search, queried on demand instead of loading catalogs into the agent’s context.
-
Graph CLI. Optional code intelligence for analysis and migration.
In the agent-run workflow the prompt is the product: generated instructions own the integration behaviour, and the compiled CLI supplies initialization, configuration, helper commands and the graph. The Ship workflow is the deliberate exception: its controller (stages, run state, oversight, evidence, guarded publication) is compiled code, and /camel-ship only delegates to it. See Architecture overview.
Where to go next
-
Install and first project installs the standalone CLI or the Camel JBang plugin and creates a first project.
-
Greenfield workflow and Migration workflow walk through the two workflows end to end.
-
Ship workflow (Technology Preview) describes the Technology Preview controller-owned workflow.
-
Source code and issue tracker: apache/camel-kit.