Orchestration

List the date and external integration tools a spawned worker can call

Form a team of execution, command and review agents with shared memory for projects

Simulate a multi-agent meeting debate that ends in decisions, arguments and risks

Analyze tricky cross-domain questions from several expert views with claim checks

Run parallel expert analyses with fact checks, then rewrite them as a first-person essay

Analyze hard questions with parallel experts, red-team review and a plain-language essay

Settle hard questions through red and blue team debate and a plain-language verdict

Stress-test a contested question with opposing red and blue team analyses

Split two or more independent tasks across parallel agents with no shared state

Run an implementation plan with independent subagent tasks in the current session

Run agency-orchestrator YAML workflows in your current session without an API key

Create installable expert teams with a leader, member agents and shared skills

Set up measured program evolution tasks where repeated search improves a program score

Create multi-role Swarm Skills with executable SwarmFlow orchestration scripts

Design and operate apps whose control loop runs on OpenRig agent teams

Decide whether agent judgment or plain code should own a workflow control loop

Address a remote OpenRig host and read the results of cross-host commands

Decide whether to do a task yourself, hand it to a subagent or to a peer agent

Coordinate a development pod through implementation, QA and design handoffs

Route agent decisions to people through durable queues and decision records

Send human decision requests and updates from an agent and resume the right work

Design multi-agent team topologies and write rig and agent spec files

Set up a continuing agent team that does reviewed work on a real repository

Coordinate assignments, review gates and blocked work across an agent rig

Run drift detectors that watch other agent rigs and decide when to intervene

Hand off unfinished agent work through durable queue items between turns

Retire an agent seat deliberately and pass it to a successor with a handover packet

Create, start, stop, restore and destroy agent rigs and verify their lifecycle state

Replace an agent seat occupant and keep clear continuity and provenance records

Grow, shrink and rebind seats in a live agent rig without breaking its wiring

Configure watchdog policies that wake or refocus stalled agents at the right level

Pass bounded, graph-informed code slices to smaller subagents to save context

Build async Python apps with agent-squad orchestrators, routers and agent types

Build on-device multi-agent Swift apps with AgentSquad routing, tools and storage

Build Node.js and TypeScript multi-agent apps with agent-squad routing and storage

Fall back to model knowledge when cascading data lookup tools keep failing

Run Python through the shell when the sandbox execution tool fails

Keep running Python reliably when the sandbox fails repeatedly

Recover from code execution failures using fallbacks and fixed workspace paths

Run Python with a fallback workflow when the sandbox fails repeatedly

Hand coding, DevOps and research tasks to an autonomous worker with shared skills

Treat a file creation task as done once the deliverable is written

Identify and classify failures that happen before a run starts

Find and handle failures that occur before any tool use or iteration begins

Fall back to shell execution when the e2b code sandbox is unavailable

Run code through a file write and shell when sandbox execution fails

Create files reliably through a fallback path when code tools fail

Run file operations through the shell when sandbox or shell tools fail

Run Python reliably when the sandbox and shell tools both fail