How to evaluate a residential browser API.

A residential browser API joins remote browser execution. It also adds managed network routing. Both services may share one vendor.

That category differs from a proxy API. A proxy API supplies network access. Your team still owns browser execution.

Managed browsers change process ownership and cleanup. Residential routing changes location and network identity. Neither contract proves the other.

Start with the job being purchased. Then require every supporting contract. Record missing behavior as unknown.

Datacenter addresses come from server infrastructure. Residential routes use participating devices in a provider's documented network. These labels describe egress, not browser execution.

Datacenter networks often prioritize speed and cost. Residential networks add peer availability questions. Sourcing and access policy also matter. So do rotation and location precision.

Neither network class guarantees task success. Require a documented selection field. Verify fallback and session continuity. Review the acceptable-use model.

Measure both classes with one authorized workload. Keep browser version, actions, geography, and timing constant. Record failures instead of hiding them in an average.

A complete service exposes three control planes. Each plane answers different engineering questions. One endpoint may cover several planes.

Lifecycle controls provision and terminate browser resources. Browser protocols carry automation commands. Network controls choose outbound routing behavior.

Teams often collapse these layers mentally. That shortcut hides ownership and failure boundaries. Separate them before evaluating any provider.

Official providers expose routing through different surfaces. Similar labels hide different operational contracts. Normalize the fields before comparing products.

The matrix records published behavior only. It awards no reliability or performance credit. Private contract terms remain outside this review.

No row proves universal task success. Test the chosen contract against real workloads. Recheck every source before procurement.

Residential routing still leaves important choices unresolved. Geography can be exact or approximate. Session labels may preserve different things.

Fallback behavior changes the user-visible result. Silent fallback favors completion over exact location. Failure preserves location requirements but stops work.

Write required behavior before selecting fields. Then test every disputed assumption. Sales language cannot replace interface documentation.

Browser cost and network cost often differ. One provider meters transferred data. Another combines subscriptions and proxy allowances.

Minimums can dominate short sessions. Media requests can dominate transferred bytes. Restricted targets can force different products.

Compare one representative workflow end to end. Include retries, idle time, and failed starts. Keep published prices outside durable prose.

Documentation starts the evaluation. Controlled tests finish it. Use accounts and targets you may access.

Run one production-shaped task. Keep the target and actions constant. Change only the provider contract.

Capture resource identifiers and timestamps. Record selected geography and observed egress. Preserve errors without exposing credentials.

Legba publishes a proposed lifecycle contract. It publishes no concrete public browser API host. It publishes no public key-issuer URL.

The preview contains no network-selection field. It contains no geography identifiers or routing fallback contract.

No Playwright endpoint is documented. No CDP endpoint is documented. No callable browser-control contract is published.

Therefore, no Legba routing quickstart is supportable. Recheck canonical docs before implementation.

Choose the smallest contract meeting actual requirements. More routing options create more failure modes. Unused flexibility still costs review time.

Rank mandatory answers before comparing vendors. Keep optional preferences separate. Reject any mandatory undocumented behavior.

Use the matrix during technical review. Use the checklist during trials. Keep results beside their source dates.

Legba should remain outside implementation comparisons today. Its preview lacks the required routing contract. Update this conclusion after documentation changes.