Run and check on your Anthropic Managed Agent workers

You have purpose-built Managed Agents for code review, tests, tech debt, and builds. Zero becomes your fleet manager — launch them, check status, fetch their output, all from Slack.

Zero connects:

Why a fleet of Worker Agents is hard to manage alone

You have invested in building specialized Worker Agents on Anthropic Managed Agents — a Tech Debt Cleaner, a PR Reviewer, a Test Runner, a Build Watcher. Each has its own system prompt, tools, and isolated cloud environment. Running any one of them today means opening the Anthropic console, creating a session, refreshing for status, and copying the output back into your workflow. When several are running at once, you lose track of who is busy, who finished, and whose output you already acted on. You tell Zero to check the fleet. It lists your agents, reports who is active, and streams results back into Slack when a session completes.

How to ask Zero to run or monitor your Managed Agents

@Zero launch my Tech Debt Cleaner agent on vm0-ai/vm0 main. Post the findings back here when the session completes.

try it

How Zero orchestrates your Managed Agent fleet

Zero reads your worker fleet

Zero calls the Anthropic Managed Agents API to list every agent you have configured, along with its description, model, and tool access. This becomes the directory of workers Zero can operate.

Zero checks session status across the fleet

For each agent, Zero queries active and recent sessions. It reports which workers are idle, which are running, how long each has been active, and what the latest output line is. No console hopping.

Zero launches and monitors on request

When you ask Zero to run an agent, it creates a new session with the target you specify (repo, PR, branch). Zero polls to completion, pulls the final output, and posts a structured summary back into Slack.

Launch specific agents, schedule recurring runs, or chain workers

Launch a specific worker

Kick off a single agent on a specific target

@Zero launch my PR Reviewer on PR #9711. Post the review comments here when done.

try it

Schedule recurring worker runs

Run a worker on a fixed cadence

@Zero every night at 10pm, launch the Test Runner on main. Post the full results here by 9am.

try it

Chain workers into a pipeline

Have one worker's output trigger the next

@Zero when the Tech Debt Cleaner finishes its next run, launch the PR Reviewer on the branch it opens.

try it

Required integrations: Anthropic Managed Agents and GitHub

Anthropic Managed Agents

Anthropic Managed Agents — Zero lists your agents, launches sessions, checks status, and fetches output. Connect it with an Anthropic API key from app.vm0.ai/connectors.

Required

GitHub

GitHub — most worker agents need repo access to run reviews, tests, or builds. Zero uses your GitHub connection to target the right branch or PR.

Required

Best practices for running a Managed Agent worker fleet

●Give each agent a distinct, human-readable name in the Anthropic console. 'Tech Debt Cleaner' reads far better in Zero's replies than an agent_011 ID.

●Keep each agent single-purpose. A worker that reviews PRs and also runs tests will drift on both. Narrow system prompts produce sharper output and easier orchestration.

●Chain workers with conditional prompts — 'when X finishes, launch Y on the output'. A fleet of focused agents becomes a pipeline without any extra infrastructure.