HarnessMatch
Menu

Agent platform

stagewise

Desktop agentic IDE for browser-aware coding and parallel worktrees.

Status
Active
License
AGPL-3.0
Evidence
Documented + code-verifiable, 27 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

An open-source Electron desktop IDE that combines a full Chromium browser, multi-provider and local models, managed project context, granular diff review, and user-orchestrated parallel agent sessions.

Good choice if

  • Vibe coders who want the browser, DOM context, app preview, and code changes in one desktop workspace
  • Developers running independent agent sessions in parallel, including isolated Git worktrees
  • Teams that want cloud subscriptions, direct API providers, or local inference behind the same visual workflow

Check before choosing

  • Connected workspaces and shell commands use the host filesystem with the signed-in user's permissions; Git worktrees and the restricted JavaScript worker are not an OS or container security sandbox
  • The shipped desktop app defaults to asking before tool calls and offers smart or always-allow modes, but its private smoke-test CLI forces always-allow and is not a supported public headless interface
  • Parallel work means separate top-level agent sessions that the user coordinates; the reviewed product registers no user-facing child-agent delegation tool even though the agent-core package contains generic child-agent primitives
See 3 more considerations
  • No shipped MCP integration was found in the product docs, dependencies, or tool registry; it appears only as a future specification item
  • The advertised 87.6% average cache-hit rate has no published dataset, model mix, version pin, or reproducible method, and no independent Stagewise harness benchmark or direct scientific evaluation was found
  • The repository contains engineering tests but only a copywriting-skill evaluation asset; CONTRIBUTING says external code pull requests are not accepted even though the website invites contributions

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.

    EvidenceHow agents work

  • Repository tool execution

    Documented

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

    EvidenceHow agents work

  • Task-aware context management

    Documented

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

    EvidenceAgent context

  • Model-independent runtime control

    Documented

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

    EvidenceTool approval modes at inspected commit

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 & isolationWorkspace isolationDocumented mechanism, not a performance score.
  • Tooling & integrationsBuilt-in 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 & permissionsApproval promptsDocumented 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

The repository has 150 engineering test files, but the only eval asset found targets a copywriting skill rather than coding-harness performance; no independent benchmark is imported. CONTRIBUTING exists but says external code pull requests are not accepted, and some architecture notes lag the current tree.

Inspect commit cb38225c2b0de27e85c10f26ed46123f487fb6f8, 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)
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
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
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
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
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Undo file changes
DocumentedProduct-supported file or session rollbackSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.

Primary evidence

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

Product record checked 2026-07-27
stagewise product overviewCurrent product positioning, desktop IDE workflow, provider choice, parallel agents, and the unsupported cache-hit marketing claimOfficial docsInstall stagewiseNative Electron desktop distribution for macOS, Windows, and Linux rather than a browser-hosted productOfficial docsHow agents workAgent loop, file and shell tools, approval behavior, and independent agents running concurrently with separate chats and modelsOfficial docsBrowser and agentEmbedded Chromium, DOM context, screenshots, navigation, tabs, page interaction, and browser-aware codingOfficial docsDiff reviewFull edit history, inline diffs, per-hunk accept and reject controls, tool-call undo, redo, and design previewsOfficial docsAgent contextWORKSPACE.md, skills, optional AGENTS.md, mentions, DOM context, context compression, and environment updatesOfficial docsModels and providersStagewise Account models, imported provider subscriptions, direct API keys, custom endpoints, and local modelsOfficial docsCustom providersOpenAI-compatible, Responses, Anthropic, Azure, Bedrock, Vertex, Ollama, LM Studio, vLLM, and self-hosted endpointsOfficial docs
View 19 additional sources
Additional first-party evidence19 sources
BYOK setupDirect provider credentials and imported Claude, GPT, Gemini, Kimi, Qwen, DeepSeek, GLM, and MiniMax accessOfficial docsWorkspacesMultiple project folders, Git worktree choices, and the explicit warning that connected workspaces are not sandboxedOfficial docsSkills and pluginsSKILL.md extensions and built-in product plugins, distinct from the unconfirmed Model Context Protocol capabilityOfficial docsStagewise 1.25.0 releaseLatest stable version, publication date, packaged desktop artifacts, and current release provenanceOfficial repositoryRepository overview at inspected commitOpen-source desktop architecture, development status, supported platforms, and repository organization at the inspected refOfficial repositoryAGPL license at inspected commitAGPL-3.0 licensing terms for the reviewed open-source repositoryOfficial repositorySecurity policy at inspected commitPrivate vulnerability reporting process and the repository security-policy artifactOfficial repositoryContribution policy at inspected commitContributor documentation and the restriction against external feature or code pull requestsOfficial repositoryMonorepo CI at inspected commitLint, build, unit-test, cross-platform, and Electron packaging checks in the public repositoryOfficial repositoryTool approval modes at inspected commitAlways-ask, always-allow, and smart approval modes, including always-ask as the persisted defaultOfficial repositoryHost shell tool at inspected commitShell command execution through the local host shell service rather than an operating-system sandboxOfficial repositoryJavaScript worker at inspected commitElectron utility-process and node:vm boundary used for the restricted JavaScript execution toolOfficial repositoryRestricted filesystem wrapper at inspected commitMount-scoped filesystem checks for the JavaScript worker, not isolation for ordinary shell executionOfficial repositoryDiff history service at inspected commitCode-level diff history primitives supporting review, attribution, undo, and redo of agent editsOfficial repositoryProduct chat agent at inspected commitThe single registered chat-agent type and its absent finish schema, limiting productized child-agent delegationOfficial repositoryGeneric child-agent primitive at inspected commitLow-level synchronous and asynchronous child-agent plumbing that is not exposed as a registered product toolOfficial repositoryInternal smoke-test CLI at inspected commitPrivate minimal Anthropic-only headless smoke utility and its forced always-allow approval postureOfficial repositoryInternal CLI package manifest at inspected commitPrivate version-zero package metadata showing that the smoke CLI is not a public product interfaceOfficial repositoryRepository evaluation asset at inspected commitThe narrow copywriting-skill evaluation file, which is not a coding-harness benchmark or independent product evaluationOfficial repository

Ecosystem discovery

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