Orchestration

Build stateful multi-actor agent graphs with LangGraph checkpoints

Orchestrate many agents across engineering, QA and DevOps from a PRD

Run independent tasks through parallel agents with different expertise

Run a multi-persona debate to stress-test architecture, design and strategy choices

Claim workspace slots before editing and coordinate delegated agent teammates

Detect the type of incoming content and route it to the right ingestion skill

Submit background jobs, spawn subagents and run parallel fan-out with progress checks

Sign tamper-evident receipts for consequential agent actions so they can be verified later

Pick a multi-agent pattern and generate tool schemas for an agent system

Turn a folder of skills into a bounded agent loop with machine-run task checks

Define how C-suite agents query each other, avoid loops and isolate their work

Design multi-agent pipelines with handoff contracts and failure handling

Run parallel agents on one task in git worktrees and merge the best result

Resume a paused autoresearch experiment from its branch and results history

Run one autoresearch iteration that keeps or discards a file edit

Show autoresearch experiment results, active loops and progress

Optimize any file against a measurable metric in an autonomous loop

Post and read coordination messages between competing agents

Run a six-phase multi-agent board deliberation to reach a strategic decision

Run a multi-role C-suite deliberation and get a board memo for a cross-team decision

Run a 13-agent C-suite team with boardroom, review and decision slash commands

Route founder questions to the right executive advisors and run multi-role boards

Have three AI models cross-review a board memo before it reaches the boardroom

Score and rank parallel agent results by metric or LLM judge

Route any founder question to the right C-role advisor or to a boardroom session

Delegate grunt coding work to cheap parallel workers to cut token costs

Start a multi-agent competition with a task and evaluation criteria

Run a full multi-agent competition from setup to merge in one command

Show progress, DAG state and branches for competing agents

Gather human review feedback in an HTML page and block completion on open blockers

Schedule autoresearch experiments hourly, daily or weekly on a timer

Find, audit and design repeatable agent loops with triggers and stop conditions

Merge the winning agent branch and clean up losing worktrees

Schedule a graded agent with cron, event triggers or on-demand runs

Set up an autoresearch experiment with target file, metric and eval command

Launch parallel agents in isolated git worktrees to compete on a task

Write deterministic JavaScript workflows that orchestrate sub-agents with loops

Write clear briefs for subagents and forks, with what to delegate and what not to

Run one operation across many files in parallel with chunked worker sessions

Plan many-file transforms and submit them to the discounted asynchronous Batch API

Coordinate a small team of read-only and writer sessions working on shared tasks

Run a prompt repeatedly on a fixed schedule or self-paced wakeups

Write Workflow tool scripts with agent options, parallel steps and resume support

Create and update reusable workflow JavaScript files for Dynamic Workflows

Manage Paseo hosts, projects, workspaces, agents, schedules and heartbeats

Get a second opinion from a separate agent on your current task

Form a two-agent committee to find root causes and plan when stuck

Hand the current task to another agent with its full context attached

Convert a PRD into the prd.json task format for an autonomous agent loop