Skip to content
Agentic Levels
Start hereLevelsTracksPromptsAbout
Get my levelGet my level

  • Start hereYour first useful task with AI.
  • Level quiz8 questions. No AI needed.
  • Fluency ScoreYour Level and a score, from your own AI.

  • LevelsThe 11 levels, in order.
  • LessonsShort lessons, filter by level.
  • TracksGuided paths by tool, role or goal.
  • Set up your AI tool
  • Starter packFree files that help AI remember.

  • PromptsReady prompts to copy and adapt.
  • ChatGPT vs Claude vs GeminiPick the right one for the job.
  • How-tos
  • ReferenceCommands for Claude Code, Codex, Antigravity.
  • ResourcesHand-picked reading and tools.

  • Your progress
  • Your next moveAn interview that picks your next skill.
  • Your next lessonsThe next lessons at your level.
  • Updates

  • Pricing
  • About
  • FAQ & Feedback
  • Work with us
  • How we handle your data
Sign inPreferences
AboutPricingHow-tosPromptsStarter packMy stackHow we handle your data

© 2026 Fuentes Studio

PrivacyTermsRefundsAccessibilityAI content policyCredits
yourCouncil
Ready to help
✦

What do you want to understand?

Ask anything about what you're learning.

←How-to guides
  1. Resources
  2. ›How-to guides

How to use Dispatch

Claude Code skill that fans tasks out to parallel fresh-context workers, 10x effective context, non-blocking, with workers that surface questions without losing state.

Levels 5 to 7On GitHub

A single Claude Code session accumulates context fast. Dispatch makes your main session a lightweight orchestrator: fan tasks out to parallel background workers, each with a full fresh 200k-token context window. Workers ask clarifying questions that surface back to you without polluting the main session context. Non-blocking, the main session is free immediately after dispatch. Supports multiple AI models per task (Claude, GPT, Gemini) with auto-discovery. Requires Claude Code as host; Cursor and Codex work as workers only.

Before you start

API keys for the models you want to route to • Python 3.10+

  1. 1Define your agent roster and routing rules
  2. 2Run and tune routing accuracy
  3. 3Review worker outputs and integrate results

Steps

01

Define your agent roster and routing rules

Dispatch works by matching task descriptions to specialized agents. Define each agent's specialty clearly, the routing quality is only as good as your agent descriptions.

Design your dispatch table
I want to set up a Dispatch routing layer for these agent types: [AGENT_LIST]. Write the dispatch configuration and routing logic. For ambiguous tasks, which agent should be the fallback? Walk me through any edge cases where routing could go wrong.
[AGENT_LIST]Your agents and specialties e.g. code-agent (implementation tasks), research-agent (information gathering), writer-agent (docs and copy)
02

Run and tune routing accuracy

Send 10-20 real tasks through the router and check which agent they land on. Adjust agent descriptions when routing is wrong, the descriptions are your routing rules.

Build a routing test set
I have these Dispatch agents: [AGENT_LIST]. Write a routing test: 15 example tasks that cover my full task space, the expected agent for each, and a simple eval script that sends each task through Dispatch and prints correct/incorrect routing with the actual vs expected agent.
[AGENT_LIST]Your agents and their specialty descriptions e.g. code-agent (Python/TypeScript implementation), research-agent (information gathering and synthesis), writer-agent (docs, copy, and explanations)
03

Review worker outputs and integrate results

Dispatched workers run in the background with fresh context windows. When a worker finishes, review its output and any clarifying questions it surfaced. Dispatch is non-blocking, your main session is free during worker execution, so use that time to dispatch more tasks or continue other work.

Review and integrate a worker's output
A Dispatch worker just completed [TASK_DESCRIPTION] and returned [WORKER_OUTPUT]. Review the output: (1) does it meet the acceptance criteria [CRITERIA]? (2) are there any clarifying questions from the worker I need to answer? (3) what integration steps are needed to wire this output into the main codebase?
[TASK_DESCRIPTION]What the worker was tasked with e.g. implement the UserProfile component with TypeScript props and Tailwind styles
[WORKER_OUTPUT]The worker's output e.g. the component file content
[CRITERIA]Acceptance criteria e.g. renders name and avatar, handles missing avatar gracefully, matches the Figma spec

How do you know it worked?

  • Tasks route to the correct agent >80% of the time on your test set
  • Fallback agent handles ambiguous tasks gracefully

If something looks off

Profile routing latency
Routing adds a round-trip, profile the overhead before committing to the pattern in production.
Cover every task type
Over-specializing agents can leave gaps; make sure every task type you'll send has a clear home agent.

Skills & Commands Reference

Dispatch's core concepts: the dispatch call, agent definitions, routing table, and worker lifecycle.

Command / SkillWhat it doesWhen to use it
dispatch(task)Fan a task out to a background worker with a fresh 200k-token context window, non-blockingAny task that doesn't need main session context, most implementation tasks qualify
Agent definitionNamed agent with a description and model assignment, description drives LLM-based routingDefine all agents before dispatching any tasks, routing quality = description quality
Routing tableMaps task types to agents, evaluated by the router LLM on every dispatch callTune when routing accuracy drops below 80% on your test set, sharpen agent descriptions
Worker questionsClarifying questions a worker surfaces back to the main session, answered without polluting main contextReview before accepting any worker output, unanswered questions mean incomplete results
Multi-model routingadvancedAssign different models (Claude, GPT, Gemini) per agent for cost or capability optimizationRoute cheap classification tasks to a smaller model, complex reasoning to Claude

See also

  • LiteLLM: unified LLM API
  • AutoGen: multi-agent conversations

What you'll do, 3 steps

  1. 1Define your agent roster and routing rules
  2. 2Run and tune routing accuracy
  3. 3Review worker outputs and integrate results

Was this helpful?