Coding agent

OpenCode

Model-agnostic coding agent with a broad, configurable surface.

Status
Active
License
MIT
Evidence
Documented + code-verifiable, 28 sources
Product record checked
2026-08-22

At a glance

An open-source, client-server coding agent across terminal, desktop, IDE, local web, SDK, and headless modes, with broad provider and subscription support, local models, granular permissions, MCP, parallel subagents, Git-backed undo and redo, and GitHub automation.

Good choice if

  • Developers who switch between providers and subscription paths
  • Local-model and custom-endpoint workflows
  • Teams that want configurable primary agents, subagents, and permissions

Check before choosing

  • Execution is host-first with no built-in OS or container sandbox; shell commands, plugins, custom tools, and local MCP servers inherit the user's privileges
  • The default Build agent has full tool access and most permission classes default to allow; granular ask/deny rules are application policy rather than an isolation boundary, and --auto approves every non-denied request
  • Git-backed undo and redo require a Git repository, omit ignored files and untracked files over 2 MB from snapshots, and cannot reverse network, process, database, or other external side effects
See 3 more considerations
  • The local web/server surface has no password unless OPENCODE_SERVER_PASSWORD is set; localhost binding is safe for ordinary local use, but network exposure needs deliberate authentication and CORS configuration
  • Browser automation is not built in: web fetch and web search retrieve content, while interactive browsing needs an external MCP or custom tool
  • Provider breadth improves choice but adds model, credential, routing, and privacy decisions; optional public sharing uploads the full conversation and remains accessible until unshared

Capability support

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

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Reusable skills
DocumentedProduct-supported reusable skill packagesSource · checked 2026-07-30Support does not establish portability, package quality, safety, or adoption.
Local models
DocumentedLocal or self-hosted model pathSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
DocumentedNon-interactive or automation surfaceSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Browser control
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Isolated execution
No built-in supportHost execution; permission rules are not process isolationSource · checked 2026-07-30The official security policy explicitly states that OpenCode has no sandbox.
Undo file changes
DocumentedProduct-supported file or session rollbackSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.

Getting started

Install OpenCode, connect one or more providers, choose a model, and tune permissions in `opencode.json`.

Open official documentation

Classification and operating model

Category fit and technical mechanisms are evidence records, not product-quality scores.

Category fit
Qualifies, 4/4 criteria
Operating model
7/7 layers documented
Inspect category criteria and operating mechanismsFirst-party records

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.

  • Repository tool execution

    Documented

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

  • Task-aware context management

    Documented

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

  • Model-independent runtime control

    Documented

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

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 & isolationHost processDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible toolsDocumented mechanism, not a performance score.
  • Context & stateManaged contextDocumented mechanism, not a performance score.
  • Lifecycle & recoveryCheckpoint/rewindDocumented mechanism, not a performance score.
  • ObservabilityLogs/transcriptsDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsPolicy controlsDocumented mechanism, not a performance score.

Measured and public context

Configuration-specific measurements and source-native activity stay separate from general product capability.

Inspect code audit, measured configurations, and ecosystem signalsContext, not a product score

Public code audit

4/5public artifacts present
Security policy
Present at inspected commit
CI workflow
Present at inspected commit
Automated tests
Present at inspected commit
Evaluation assets
Not found
Contributor documentation
Present at inspected commit

The 1.18.5 tree contains 901 test-like files and 26 workflows. Two UI performance benchmark paths measure application rendering, not coding-task success; no independent or complete model × harness evaluation record is inferred from them.

Inspect commit e5cc278dec9294a627a7b05f47ce6a564408c1a2, 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
Context, not quality

Public ecosystem signals

Source-native observations for exact mapped artifacts and reviewed stable release trains. Different units and populations stay separate, and missing coverage is never treated as zero.

View this harness in Usage
  • Latest stable releasev1.18.29Released 2026-09-04; 47 stable releases in 90 daysOpen release
  • Homebrew 30d events28.84KFormula: opencodeOpen artifact
  • npm last-month downloads9.83MPackage: opencode-aiOpen artifact
  • Release asset downloads17.76M865 stable releases; 7050 matched assetsOpen artifact
  • GitHub stars205.78KFull-source repository; 26.85K forksOpen artifact

Routing, package retrievals, release downloads, editor installs, and repository interest observe different populations. They are never added together and never affect capability evidence, classification, or measured results.

Interpretation rulesSignals checked

First-party evidence

Each capability claim links to the first-party record that supports it.

28 first-party sourcesProduct record checked

Product and interfaces

6 sources
View 5 more sources

Execution and control

3 sources
View 2 more sources

Agents, state and recovery

7 sources
View 6 more sources

Automation and extensions

6 sources
View 5 more sources

Enterprise and operations

3 sources
View 2 more sources

Releases and public code audit

3 sources
View 2 more sources

Ecosystem context

OpenCode model-usage data First-party live telemetry for OpenCode Go and Zen model usage, daily unique users, session and token cost, cache ratio, provider share, and geography. It is product-operated usage context, not a benchmark or evidence of task success, and does not affect capability claims or classifications. Observed 2026-07-28.