HarnessMatch
Menu

Extensible harness

Amp

Opinionated hosted agent for local work, durable cloud orbs, and scheduled automation.

Status
Active
License
Proprietary
Evidence
Documented, 19 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

A proprietary multi-model coding agent with terminal, editor, and web control, non-interactive and runner modes, durable cloud orbs, schedules, MCP, skills, plugins, and built-in or programmable subagents.

Good choice if

  • Hands-off vibe coding in disposable managed cloud orbs
  • Long-running, scheduled, or remotely controlled agent work
  • Teams extending one hosted thread system with MCP, skills, plugins, SDK calls, and delegated agents

Check before choosing

  • Local tools run with host privileges and without approval by default; permission plugins or managed policies must be configured explicitly
  • Orbs are paid managed cloud sandboxes, while the SDK's orb executor ignores local-only permission and tool settings in favor of project configuration
  • No harness checkpoint or file rollback was verified; diff review, Git, and durable thread resume are different mechanisms
See 2 more considerations
  • The service is proprietary and not self-hostable, with no documented local open-weight model path
  • Workspace thread visibility and administrator access require deliberate governance for sensitive code and prompts

Why it qualifies as a coding harness

This confirms category fit, not product quality. Every required criterion links back to first-party evidence.

Qualifies4 of 4 required criteria evidenced

  • Adaptive agent loop

    Documented

    The system repeatedly observes results and chooses the next action instead of following a fixed one-pass graph.

    EvidenceAmp owner’s manual

  • Repository tool execution

    Documented

    The system can use tools to inspect and change a repository or its execution environment.

    EvidenceAmp owner’s manual

  • Task-aware context management

    Documented

    The runtime assembles, updates, compacts, retrieves, or persists task-relevant context while work proceeds.

    EvidenceScheduled agents

  • Model-independent runtime control

    Documented

    Permissions, budgets, interruption, policy, or stop controls operate outside the model's own text generation.

    EvidencePermissions and plugins

Membership establishes category fit only. It does not score quality, safety, autonomy, model capability, or benchmark performance. · Read the membership rule.

How it works under the hood

Seven mechanisms mapped from first-party records. These labels describe what the harness provides, not how intelligent its model is.

7/7layers documented
  • Execution & isolationSandbox availableDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible + browserDocumented mechanism, not a performance score.
  • Context & statePersistent stateDocumented mechanism, not a performance score.
  • Lifecycle & recoveryManaged recoveryDocumented mechanism, not a performance score.
  • ObservabilityStructured tracesDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsHost accessDocumented mechanism, not a performance score.

Public code audit

Unrankedsupport-only repository
Security policy
Not found
CI workflow
Present at inspected commit
Automated tests
Not found
Evaluation assets
Not found
Contributor documentation
Not found

The public repository is an older supplemental collection of examples and workflow guides with a Markdown lint workflow; it does not expose the proprietary Amp client, durable execution service, orb runtime, or product test suite.

Inspect commit 435ffc0b4d82dfcb083848088411afb44b1ff55f, checked 2026-07-27

Measured configurations

No benchmark run passes the full metadata admission policy for this harness yet. Missing data is not scored as zero.

Benchmark policy and all runs

Capability support

Documented first-class product support, checked against the sources below.

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Local models
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
DocumentedNon-interactive or automation surfaceSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Browser control
DocumentedBrowser-capable local or managed agent workflowSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Isolated execution
Depends on surfaceManaged cloud orbs; local tools run on the hostSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Undo file changes
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.

Primary evidence

Each capability claim is tied to a first-party record and a verification date.

Product record checked 2026-07-27
Amp owner’s manualCurrent product surface, terminal and editor interfaces, thread model, modes, built-in tools, and supported platformsOfficial docsProjects and changes workflowRepository-linked settings and secrets plus ship, branch, pull-request, and custom changes workflowsOfficial docsCloud orbsManaged cloud execution environment, lifecycle, setup and resume scripts, portals, secrets, and billingOfficial docsRemote runnersLocal or remote runner registration and dispatch to user-controlled machinesOfficial docsSubagents and reviewAutomatic isolated-context subagents, delegation limits, and separate review agentsOfficial docsMCP and workspace trustLocal and remote MCP, skill-scoped tools, workspace approval, OAuth, loading order, and server policyOfficial docsPermissions and pluginsNo-approval default and optional allow, reject, ask, or delegated rules through plugins and managed settingsOfficial docsScheduled agentsOne-time and recurring schedules that wake a thread with its saved prompt, context, and historyOfficial docs
View 11 additional sources
Additional first-party evidence11 sources
Thread sharing and governancePrivate, group, workspace, and unlisted visibility plus administrator access boundariesOfficial docsCLI execute modeInteractive CLI, non-interactive execute mode, orb execution, streaming JSON, and remote runner modeOfficial docsPlugin APITool and lifecycle hooks, custom agents and threads, executor selection, background work, and permission decisionsOfficial docsTypeScript SDKProgrammatic local and orb execution, streams, structured options, permission rules, and orb-setting limitationsOfficial docsSecurity referenceService architecture, retention, thread audit trail, prompt-injection controls, secret redaction, and enterprise controlsOfficial docsDurable agent foundationDistributed durable execution and unified monitoring across web, mobile, editor, and CLIOfficial announcementAgents in Orbs announcementFresh managed machines, remote inspection, change synchronization, and non-interactive orb launchOfficial announcementRemote runners announcementRemote thread creation and headless runner mode on user-controlled machinesOfficial announcementSchedules announcementRecurring agent wakeups with saved prompt, context, history, and downstream integrationsOfficial announcementCustom agents announcementPlugin-defined main agents, subagents, tool pipelines, background threads, and parallel custom workersOfficial announcementSubscription and orb limitsSubscription tiers, included orb hours, mode access, linked model subscriptions, and pay-as-you-go continuationOfficial announcement