Coding agent

Reasonix

Multi-provider coding harness with durable context, delegated agents, and policy controls.

Status
Active
License
MIT
Evidence
Documented + code-verifiable, 17 sources
Product record checked
2026-08-02

At a glance

An MIT-licensed coding harness whose Rust runtime implements an adaptive model-and-tool loop, repository and shell mutation, automatic context compaction, durable project memory, delegated agents, MCP, skills, policy rules, OS shell sandboxing on supported platforms, persistent checkpoints, and structured headless execution. The same core is exposed through terminal, desktop, local web, editor, and automation surfaces.

Good choice if

  • Multi-provider or self-hosted endpoint workflows that need one harness across terminal, desktop, web, and editor clients
  • Long repository tasks that benefit from automatic compaction, durable project memory, resumable goals, and delegated subagents
  • Local macOS or Linux execution that needs explicit permission policy, OS shell sandboxing, structured events, and file checkpoints

Check before choosing

  • The cache-oriented design is documented, but the project's cost and cache-hit marketing figures are not admitted as independent product measurements
  • The OS sandbox is enabled by default for Bash on macOS and supported Linux installations, but Windows shell commands run without that isolation; network access is allowed by default, and users can disable the sandbox
  • Workspace file tools remain path-confined, but installed MCP servers, hooks, plugins, language servers, and other trusted extensions have separate execution boundaries; MCP tool calls do not receive Reasonix's per-tool approval flow
See 5 more considerations
  • Checkpoints cover previewable Reasonix file-edit tools and conversation state, not Bash, move operations, remote services, or other external side effects; rewind can overwrite file changes made outside Reasonix after the snapshot
  • Headless writes require an explicit automatic permission mode; deny rules still apply, but unattended use removes interactive approval and needs independent time, spend, credential, and network limits
  • Self-hosted model support means connecting an OpenAI-compatible or Anthropic-compatible endpoint; Reasonix does not bundle a local model runtime
  • The local web server binds to loopback without authentication by default; remote exposure requires an explicit token or password and an appropriate network boundary
  • The built-in web tool retrieves page content rather than controlling an interactive browser; browser automation would require a separately trusted extension such as an MCP server

Capability support

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

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-08-02The source establishes the mechanism, not its quality or availability in every mode.
Reusable skills
DocumentedProduct-supported reusable skill packagesSource · checked 2026-08-02Support does not establish portability, package quality, safety, or adoption.
Local models
DocumentedSelf-hosted OpenAI-compatible or Anthropic-compatible endpointsSource · checked 2026-08-02Reasonix does not bundle a local model runtime; endpoint availability, model capability, and isolation remain external.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-08-02The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
DocumentedNon-interactive or automation surfaceSource · checked 2026-08-02The 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
Depends on surfaceDefault OS shell sandbox on macOS and supported Linux installations, plus workspace-confined file tools across supported platformsSource · checked 2026-08-02Windows shell execution is unconfined, network access is allowed by default, the OS sandbox can be disabled, and trusted extensions retain separate boundaries.
Undo file changes
DocumentedPersistent file and conversation snapshots for previewable Reasonix edit toolsSource · checked 2026-08-02Checkpoints exclude Bash, move operations, remote services, and other external side effects; rewind can overwrite later external edits.

Getting started

Install the stable Reasonix CLI, configure a supported provider or compatible endpoint, review workspace permissions and platform-specific sandbox coverage, then enable headless auto mode, MCP servers, plugins, hooks, or remote web access only inside an independently bounded environment.

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

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 toolsDocumented mechanism, not a performance score.
  • Context & statePersistent stateDocumented mechanism, not a performance score.
  • Lifecycle & recoveryCheckpoint/rewindDocumented 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.

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

5/5public artifacts present
Security policy
Present at inspected commit
CI workflow
Present at inspected commit
Automated tests
Present at inspected commit
Evaluation assets
Present at inspected commit
Contributor documentation
Present at inspected commit

The tagged v1.19.2 tree contains 2,286 tracked paths, 26 workflows, 837 test-like paths, security and contributor documentation, and 22 benchmark- or evaluation-like paths. The evaluation assets are project-owned development tasks and tests rather than an independent immutable model × harness × environment × budget × attempts result, so no product score is imported.

Inspect commit c46e3af1c2732fe2b3dedb0bd47eb39a629357d2, checked 2026-08-02

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 releasev1.19.2Released 2026-08-02; 68 stable releases in 90 daysOpen release
  • npm last-month downloads45.33KPackage: reasonixOpen artifact
  • Release asset downloads59.03K48 stable releases; 288 matched assetsOpen artifact
  • VS Code installs2.06KExtension: SivanLiu.reasonix-agentOpen artifact
  • Open VSX downloads896Extension: SivanLiu/reasonix-agent; latest 0.3.0Open artifact
  • GitHub stars28.88KFull-source repository; 1.86K 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.

17 first-party sourcesProduct record checked

Product and interfaces

3 sources
View 2 more sources

Execution and control

3 sources
View 2 more sources

Agents, state and recovery

4 sources
View 3 more sources

Automation and extensions

3 sources
View 2 more sources

Enterprise and operations

1 source

Releases and public code audit

3 sources
View 2 more sources