HarnessMatch
Menu

Agent platform

Coder Agents

Self-hosted agent control plane with governed workspaces.

Status
Active
License
AGPL-3.0 core with separately licensed enterprise components
Evidence
Documented + code-verifiable, 20 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

A beta, self-hosted coding agent for teams already using Coder. It works from a web chat or API, can create approved workspaces, delegate parallel subagents, and use the models and tools configured by an administrator.

Good choice if

  • Teams that already run Coder and want one web interface for delegated coding work
  • Organizations that need centrally selected models, user permissions, cost controls, and inspectable activity
  • Parallel or automated jobs that should start from administrator-approved workspace templates

Check before choosing

  • Coder Agents is beta, and its Chats API is still experimental
  • It requires a self-hosted Coder deployment and administrator setup; it is not a lightweight tool for an individual project
  • Workspace isolation depends on the selected template and its network rules; broad template access is not a security sandbox by itself
See 2 more considerations
  • Graphical desktop and browser work are experimental and require a supported provider plus a desktop-enabled workspace template
  • Chat history can resume after a workspace is replaced, but there is no first-class file checkpoint or rewind feature

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.

    EvidenceCoder Agents overview

  • Repository tool execution

    Documented

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

    EvidenceBuilt-in tools

  • Task-aware context management

    Documented

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

    EvidenceCoder Agents architecture

  • Model-independent runtime control

    Documented

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

    EvidencePlatform controls

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 & isolationManaged isolationDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible + browserDocumented mechanism, not a performance score.
  • Context & stateManaged contextDocumented 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 & permissionsPolicy controlsDocumented mechanism, not a performance 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

This audit covers the broader Coder platform repository that contains Coder Agents, not a standalone agent-only codebase.

Inspect commit 6120fb598865, 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
OptionalExperimental computer-use subagent in desktop-enabled workspacesSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Isolated execution
Depends on surfaceIsolation depends on the selected workspace templateSource · 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
Coder Agents overviewBeta status, self-hosted control-plane loop, web chat, workspaces, subagents, persistence, and built-in toolsOfficial docsCoder Agents architectureControl-plane loop, workspace execution path, tool boundaries, stored chat state, identity, and network isolation conditionsOfficial docsModels and providersAdministrator-selected providers, OpenAI-compatible and self-hosted endpoints, model options, routing, credentials, and BYOKOfficial docsBuilt-in toolsVisible tool calls, user-scoped permissions, workspace creation, file operations, command execution, and tool limitsOfficial docsPlatform controlsAdministrative model and prompt policy, template allowlists, MCP policy, lifecycle, retention, and spend controlsOfficial docsMCP server controlsAdministrator-configured MCP servers, authentication, availability policies, tool allowlists, denylists, and identity headersOfficial docsWorkspace skills and MCPRepository skills and workspace MCP discovery, transports, timeouts, tool naming, and path-safety rulesOfficial docsCoder Agents getting startedVersion prerequisite, roles, setup, API automation, template selection, network caveat, human review, and beta limitationsOfficial docs
View 12 additional sources
Additional first-party evidence12 sources
Chats REST APIExperimental programmatic chat sessions, child-agent records, message state, model configuration, MCP selection, and streamingOfficial docsSpend managementPer-user and per-model cost, token, message, chat, and rolling spend-limit recordsOfficial docsChat debug loggingOptional structured traces of requests, responses, tool activity, token use, retries, errors, export, and retentionOfficial docsVirtual desktopExperimental computer-use subagent, desktop interaction, provider requirements, workspace module, and API configurationOfficial docsChat runtime architecture at inspected commitDatabase state machine, queueing, retries, worker ownership, API surface, streaming, and durable chat transitionsOfficial repositorySubagent implementation at inspected commitChild-chat creation, model override resolution, subagent messaging, interruption, and agent listing implementationOfficial repositoryWorkspace MCP implementation at inspected commitWorkspace MCP discovery, provider-safe tool names, connection routing, parallel tool metadata, and error handlingOfficial repositoryWorkspace execution implementation at inspected commitHeadless shell execution, timeouts, background process tracking, working directories, environment, and output boundsOfficial repositoryCoder Agents runtime tests at inspected commitFirst-party automated tests for chat runtime, providers, MCP, permissions, streaming, retries, and workspace behaviorOfficial repositoryCoder repository licenses at inspected commitAGPL-3.0 license for the open-source repository outside separately licensed enterprise directoriesOfficial repositoryCoder enterprise license at inspected commitSeparate license terms for code under the repository enterprise directoryOfficial repositoryCoder repository overview at inspected commitSelf-hosted platform scope, workspace infrastructure, native agent loop, model routing, identity, cost tracking, and audit claimsOfficial repository

Ecosystem discovery

OpenRouter apps Discovery signal only; usage rank is not used as a quality or capability score. Observed 2026-07-27.