Platform Setup

Operations Agent Usage Tasks Runners Workspace Chat
This board shows platform configuration status. Secrets are stored server-side and never displayed after saving. Configuration values shown as runtime_config are operator-visible settings. Values classified as dev_fixture, test_fixture, governance_constant, docs_example, or seed_data are internal and not operator-editable. See hardcode-classification.md for details.
Manage providers and model profiles. Includes OpenRouter Workspace Chat configuration. Providers and models are loaded from the D1-backed catalog. Secrets are never displayed. Seed/bootstrap entries are not editable beyond enable/disable.
Loading...
Providers
Model Profiles
Active route policies determine how providers and models are selected for each run type. Fallback chains are ordered by eligibility.
3 policies active
Template IDs: analysis-report-v1, verification-review-v1, implementation-read-only-v1, follow-up-existing-session-v1, runner-smoke-v1
V1 templates available
Allowlist: analysis, verify, execute_read_only, follow_up, smoke
V1 run types active
Runner registry, authority profile and readiness.
See Runner Dashboard /ui/runners
Readonly is the safe default and cannot execute implementation work.
Code agent requires explicit local opt-in, OpenCode readiness, target allowlist and managed worktree capability.
Start the runner using the bootstrap guidance, then verify a fresh heartbeat and capability report in /ui/runners.
Tune runner polling and heartbeat intervals to control D1 traffic. Values are exported for the local runner settings file; dashboard refresh is configured on the Runner Dashboard.
Safe bounds enforced by runner
The runner reads this file from .phlowao/runner-settings.json. Values outside the hard safety bounds are clamped locally.
Preflight for the authorized dev/pre-production Chat-to-completion flow. Every gate must be ready before a real mutation run.
Explicit target required
Required setup gates
Loading governed E2E targets…
Preflight has not run.
  1. Approved disposable dev/pre-production target
  2. Validated GitHub connection with required scoped capabilities
  3. Online code-agent runner with managed-worktree readiness
  4. Explicit sprint authorization and exact-SHA requirement
  5. Bound validation profile and protected merge policy
  6. Durable evidence, write-back and cleanup procedure
No target is configured in this environment. Live mutation remains blocked until all gates are explicitly satisfied.
Open governed task flow Check runners
Read and update the versioned execution envelope for this tenant. Only dev/pre-production is executable; production execution and auto-merge are rejected server-side. Non-mutating work uses the readonly runner profile (dry-run was removed per #1231).
Every save bumps the version
Create and select operator validation profiles: required checks, evidence TTL, baseline branch, clean-scope requirement, and post-merge checks. Branch protection and clean scope can never be disabled; saving a profile invalidates evidence collected under older configuration.
Scope and protection are immutable
Loading validation profiles…
Merge-gate status preview (exact expected SHA, evidence freshness, check status, branch protection, mergeability, scope) with typed blockers. The merge action verifies everything server-side through the GitHub connection adapter — client-asserted evidence is rejected — and requires the server-side operator approval secret (MERGE_OPERATOR_TOKEN); the canonical policy evaluator runs immediately before the merge adapter, which re-verifies the exact head SHA.
No merge without operator approval and all server-verified gates green
Gate status has not been checked.
Manage GitHub personal access tokens for repository access, issue intake, and source control integration. Tokens are validated server-side, encrypted before storage, and never returned to the browser.
Loading...
OpenCode, Codex CLI, Claude Code, Aider, Goose, etc.
Via runner discovery
Repo allowlist, branch policy, write scope.
V1 allowlist in persisted runtime definitions
Hard gates before assignment/execution.
Active (no bypass controls)
Pre-accepted permission profiles that auto-approve safe, low-risk external-directory (temp) requests without a manual Operator Action. Unsafe, broad, secret, and wrong-scope requests always fall back to manual review. The built-in default covers OS temp directories only (%TEMP%, /tmp).
Loading...
Profiles
Normalized result/evidence capture.
V1 evidence events active
Which configuration values are operator-visible runtime settings versus internal fixtures/constants. runtime_config values are editable in Setup. dev_fixture, test_fixture, governance_constant, docs_example, and seed_data are internal and not operator-editable.
See docs/hardcode-classification.md
Local/stage/prod endpoints.
See /health for runtime status
Safe health checks for runtime bindings.
Check /health