← GUI catalog

Multi-harness workspace

QM

A self-hosted web and Slack control plane for scoped, durable agent work across four named harnesses.

Status
Active
License
MIT
Source access
Open source
Record checked

At a glance

Startups that need personal and shared agent scopes, background work, policy controls, and isolated cloud workspaces in one deployment.

Good choice if

Startups that need personal and shared agent scopes, background work, policy controls, and isolated cloud workspaces in one deployment.

Check before choosing

QM is early experimental software, requires an operator-owned cloud deployment, documents material security gaps, and does not currently establish an integrated visual code-review surface or arbitrary-CLI support.

Workflow fit

Existing evidence gates classify four predefined workflows. Bands are non-numeric and do not rank product quality.

  • One agent + review

    Conditional fit

    Stay close to one coding session and inspect changes in a dedicated visual surface.

    The product may fit, but current sources do not establish visual review.
  • Parallel local tasks

    Good fit

    Run several coding tasks on one machine without their branches or files colliding.

    Every required mechanism is documented; current sources do not establish the preferred visual review mechanism.
  • Remote control

    Strong fit

    Start or revisit coding work away from the machine where the repository is running.

    Current sources document remote execution or access, shared team access.
  • Shared team workspace

    Good fit

    Let teammates enter the same live workspace to watch, steer, preview, or review work.

    Every required mechanism is documented; current sources do not establish the preferred visual review mechanism.

Harness coverage

Four named harness adapters drive the same core. Their own access, model routes, capabilities, and security posture remain independent from QM.

4 named integrations
  • Claude Code
  • Codex
  • OpenCode
  • Pi

Workflow mechanisms

Unknown means the current first-party record does not establish the mechanism; it is not converted into evidence of absence.

4 of 5 documented
  • parallel sessions

    Documented

    The official web-UI example shows concurrent sessions, while the core persists and supervises independent session runs.

  • workspace isolation

    Documented

    Each person or room receives a scoped durable sandbox, files, memory, keychain view, permissions, and background jobs.

  • visual review

    Not established

    The current first-party record does not establish an integrated diff, code-review, or merge surface.

  • remote execution or access

    Documented

    Web and Slack clients steer work in an operator-owned cloud deployment, with durable sandboxes, queues, crons, watches, and background workers.

  • shared team access

    Documented

    Personal scopes coexist with shared Slack channels, group messages, projects, shared skills, and organization policy.

Technical and public context

Implementation inspection and source-native activity provide investigation context. Neither changes workflow fit.

Inspect implementation record and public activityContext, not workflow fit

Public activity signals

Daily source snapshots provide ecosystem context only. They do not affect workflow fit or capability claims.

1 mapped source

Implementation record

Inspected at commit 7f2c916360f1.

  • The core supervises durable session runs and routes them to four named harness adapters.
  • Scope-specific state, policy, persistence, sandbox routing, background workers, and web or Slack surfaces are implemented in the pinned tree.
  • The public code establishes workflow mechanisms but not arbitrary-CLI support, visual-review completeness, security guarantees, or task-success rates.
Inspected paths
  • README.md
  • SECURITY.md
  • deployment.md
  • src/core/orchestrator.ts
  • src/harness/harness-router.ts
  • src/runs/worker.ts
  • src/sandbox/sandbox-routing.ts
  • plugins/web-ui/README.md

QM calls itself early experimental software and documents material gaps in command policy, browser approval coverage, screening, egress enforcement, governance, and data retention. The audit confirms implementation presence rather than operational safety or reliability.

First-party evidence

Sources establish specific current GUI claims; their count does not add fit points.

4 first-party sourcesGUI record checked