Start with opendevbrowser_prompting_guide or opendevbrowser_skill_load opendevbrowser-best-practices "quick start" before low-level browser commands, then load the validated capability lanes when you need the current reliable transcript plus browser replay, desktop observation, research, shopping, and browser-scoped computer-use workflows.
Choose the installation route that matches your operating model.
Best for first runs, local trials, and CI smoke checks.
Install once for repeated local workflows and shell scripts.
Use for pre-release first-run simulation from a local tarball.
Three steps to complete an end-to-end run with explicit verification.
Install runtime and extension prerequisites for the mode you need (managed, extension, or cdpConnect).
Start daemon, verify extension handshake status, then launch/connect and capture a baseline snapshot before actions.
Execute actions/workflows, then verify status, diagnostics, cookie policy results, and cleanup state for deterministic completion.
Generated onboarding metadata and generated help drive these first-contact entry points.
Use the public browser replay lane when you need temporal browser evidence before or after a fragile flow.
Use the public read-only desktop observation plane for sibling desktop evidence on macOS; availability, window inventory, and accessibility probes use the local swift command, while screenshots use macOS screencapture outside extension relay.
Control the bounded browser-scoped computer-use challenge lane with --challenge-automation-mode; the optional helper is not a desktop agent.
Keep installed CLI help output aligned with the generated docs, onboarding, and public-surface manifest.
Use both command forms during onboarding validation to confirm parity:
First-run onboarding and skill-pack guidance are mirrored from the public repo and exposed through these canonical sources.
Capture UI notes, element context, and structured payloads for implementation handoff.
Annotate live interfaces, attach comments to selected elements, and export payloads plus screenshots for reproducible review.
Run the annotate command in managed or extension mode, then validate screenshots and annotation payloads in output artifacts.
Generated categories map directly to source-of-truth docs, tool code, and skill definitions in this repository.
1 generated pages
Open category78 generated pages
Open category3 generated pages
Open category2 generated pages
Open category6 generated pages
Open category1 generated pages
Open category1 generated pages
Open category10 generated pages
Open category71 generated pages
Open category3 generated pages
Open categoryOperational friction is documented with explicit checks and recovery commands.
Start with status, handshake, and mode checks before executing actions or workflows.
Follow command-level recovery sequences to isolate auth, relay, and target failures quickly.
Workflow runs can explicitly control session cookie behavior per execution.
Use `--use-cookies=false` or policy `off` for open browsing and deal-hunting runs that do not require login.
Use policy `auto` to attempt cookie injection and continue when cookies are unavailable or rejected.
Use policy `required`; failures return explicit `auth_required` with structured `cookieDiagnostics` details.
Use docs, repository issues, and release records for support workflows.
Source code, issues, and releases for implementation truth.
Command-level behavior, flags, and examples from canonical docs.
Recovery patterns, diagnostics, and mode-specific operational checks.
