HarnessMatch
Menu

Agent platform

OpenHands

Agent platform with selectable local, sandboxed, and remote runtimes.

Status
Active
License
MIT core; separate enterprise terms
Evidence
Documented + code-verifiable, 23 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

An open agent platform and developer control center with CLI, web, SDK, and automation surfaces; provider and local-model flexibility; MCP and browser tools; and local, Docker, or remote execution environments.

Good choice if

  • Longer autonomous tasks in an isolated runtime
  • Self-hosted agent-platform and SDK workflows
  • Issue, IDE, and automation integrations beyond one terminal session

Check before choosing

  • The platform is materially heavier than a focused CLI and spans separately versioned OSS app, Cloud, and SDK release trains
  • Docker is the recommended local isolation; ProcessSandbox runs directly on the host and is explicitly unsafe for untrusted tasks
  • Headless CLI runs in always-approve mode, and direct SDK execute_tool calls bypass the normal security analyzer and confirmation policy
See 4 more considerations
  • Security analyzers and confirmation policies reduce risk but do not make model-generated actions trustworthy; remote deployments still require infrastructure hardening
  • Persistent memory is opt-in, while session persistence, condensation, and memory solve different context problems
  • Scheduled and event-driven automations run unattended in fresh sandboxes with the user's configured LLM, stored secrets, MCP integrations, and Git-provider credentials, so trigger scope and credential scope still require review
  • Local models are supported with documented functional limitations, and the project evaluation harness is not independent evidence of product quality

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.

    EvidenceOpenHands 1.11.0 source

  • Repository tool execution

    Documented

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

    EvidenceSandbox overview

  • Task-aware context management

    Documented

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

    EvidenceConversation persistence

  • Model-independent runtime control

    Documented

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

    EvidenceSecurity and action confirmation

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 & recoverySession resumeDocumented mechanism, not a performance score.
  • ObservabilityStructured tracesDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsPolicy controlsDocumented mechanism, not a performance score.

Public code audit

3/5public artifacts present
Security policy
Not found
CI workflow
Present at inspected commit
Automated tests
Present at inspected commit
Evaluation assets
Not found
Contributor documentation
Present at inspected commit

The OSS 1.11.0 application tree contains 589 test-like files, 26 workflows, and contributor documentation but no root security policy or dedicated evaluation suite. Cloud and SDK follow separate release trains, and split-licensed or hosted components are outside this root-repository audit.

Inspect commit 11ca68ab2e15dcd85c21e4d7d3409e7a259369ac, 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
DocumentedLocal or self-hosted model pathSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
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
DocumentedBuilt-in or product-supported browser controlSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Isolated execution
DocumentedDocumented execution-isolation mechanismSource · 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
OpenHands V1 documentation indexSeparation of product and SDK surfaces plus the current first-party documentation inventoryOfficial docsOpenHands 1.11.0 sourcePinned OSS application source, licensing boundary, tests, CI workflows, and contributor documentationOfficial repositorySandbox overviewDocker, unisolated local process, and remote execution choices in the V1 productOfficial docsDocker sandboxRecommended container isolation, mounts, networking, image selection, and local prerequisitesOfficial docsProcess sandboxHost-process execution, absence of container isolation, trusted-task warning, and platform limitsOfficial docsHeadless modeNon-interactive execution, always-approve behavior, output modes, scripting, and CI useOfficial docsSecurity and action confirmationConfirmation policies, action risk analyzers, user review flow, and direct-tool bypass boundaryOfficial docsTask Tool SetSynchronous delegation to specialized subagents, task isolation, results, and limitationsOfficial docs
View 15 additional sources
Product and interfaces4 sources
Browser useBrowser tool configuration, navigation and interaction capabilities, screenshots, and sandbox relationshipOfficial docsLLM subscriptionsChatGPT Plus and Pro authentication for Codex models and subscription constraintsOfficial docsLocal LLMsLocal-model configuration, GPU guidance, model requirements, and documented functional limitationsOfficial docsCLI installationSupported CLI installation methods, local requirements, startup, configuration paths, upgrades, and platform setupOfficial docs
Execution and control1 sources
Custom sandbox imagesV1 agent-server container boundary, custom image construction, local-GUI scope, fallback behavior, base images, and runtime configurationOfficial docs
Agents, state and recovery2 sources
Conversation persistenceConversation state storage, restoration, resume workflows, event persistence, and storage backendsOfficial docsPersistent memoryOpt-in cross-conversation two-tier memory, retrieval, storage, lifecycle, and privacy considerationsOfficial docs
Automation and extensions4 sources
Model Context ProtocolMCP server discovery, tool integration, configuration, authentication, and execution modelOfficial docsCLI MCP server managementHTTP, SSE, and stdio MCP server configuration, OAuth and header authentication, enablement, environment variables, and CLI managementOfficial docsScheduled automationsScheduled unattended conversations, fresh sandbox creation, stored secrets, provider credentials, MCP integrations, run history, and user scopeOfficial docsEvent-based automationsGitHub-event and custom-webhook triggers, organization routing, filters, signing-secret workflow, service accounts, and failure modesOfficial docs
Enterprise and operations2 sources
Observability and tracingOpenTelemetry spans, supported backends, instrumentation, debugging, and export configurationOfficial docsMetrics trackingToken, cost, and latency metrics plus conversation-level aggregation and exportOfficial docs
Releases and public code audit2 sources
Evaluation harnessProject-owned evaluation infrastructure and protocol surface without importing a product scoreOfficial docsOpenHands 1.11.0 releaseDated OSS product release corresponding to the pinned implementation snapshot, distinct from separately versioned cloud releasesOfficial announcement