Source: https://internal-agents.com/agents/browserbase-bb # Browserbase — bb One generalized agent in Slack that writes PRs, investigates sessions, queries the warehouse, logs feature requests, and runs browser agents across engineering, ops, sales, and support. - Approach type: Task agent - Deployment stage: Deployed - Autonomy: Drafts reviewed - Evidence strength: Detailed primary - Status: Internal - First reported year: 2026 - Work: Coding, Code review, Support, Customer success, Research - Interfaces: Slack, Web, Webhook - Invocation: Event driven, Interactive - Entry reviewed: 2026-09-09 ## Where people stay involved - **coding request → reviewed pull request** — Work product review · Level 3 ## Overview ### Summary One generalized agent in Slack that writes PRs, investigates sessions, queries the warehouse, logs feature requests, and runs browser agents across engineering, ops, sales, and support. Fact · Reported · High confidence · `browserbase-bb--summary` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ## How it works ### Supporting component Markdown playbooks lazy-loaded per task so the general agent stays small Fact · Reported · High confidence · `browserbase-bb--primitives-0` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Supporting component Scoped permissions per session limit the actions available to the agent Fact · Reported · High confidence · `browserbase-bb--primitives-1` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Supporting component The sandbox runs arbitrary code yet never touches a secret Fact · Reported · High confidence · `browserbase-bb--primitives-2` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ## Supervision evidence ### Operating model assessment Level 3 for coding request → reviewed pull request; human attention boundary: work-product-review. Inference · Catalog judgment · High confidence · `browserbase-bb--operating-models-0` Confidence reason: The source documents agent-authored pull requests while the enclosing workflow retains human review. Qualifications: - Observation date: 2026 Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ## Implementation details ### Sandbox Ephemeral Linux VM; pre-warmed snapshot rebuilt every 30 min; idles out after 30 min Fact · Reported · High confidence · `browserbase-bb--architecture-sandbox` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Harness OpenCode core loop with 6 tools (read/write/edit/exec/safebash/skill) Fact · Reported · High confidence · `browserbase-bb--architecture-harness` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Model Frontier models Fact · Reported · High confidence · `browserbase-bb--architecture-model` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Interfaces slack, web, webhook Fact · Reported · High confidence · `browserbase-bb--architecture-interfaces` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Tool access exec routes through a serverless integration proxy (Snowflake, HubSpot, Pylon, Grafana); the sandbox never sees real secrets Fact · Reported · High confidence · `browserbase-bb--architecture-tool-access` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Knowledge Key repos cloned into /knowledge/; skills (markdown) loaded on demand Fact · Reported · High confidence · `browserbase-bb--architecture-knowledge` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Credentials Credential brokering; the sandbox boots with references + rotating session tokens only; the proxy holds real creds; egress injection for a few hosts Fact · Reported · High confidence · `browserbase-bb--architecture-credentials` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ## Reported metrics ### Headline claim Feature-request pipeline at 100% coverage with zero human effort Metric · Reported · High confidence · `browserbase-bb--headline-metric` Confidence reason: A linked first-party source states the claim. Qualifications: - Reported by: Browserbase - Scope: Automatic feature-request scanning of every closed support ticket and meeting transcript - Denominator: Closed support tickets and meeting transcripts Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) · Preserved content.md, lines 26 ### Key observation Feature-request pipeline at 100% coverage, zero human effort Metric · Reported · High confidence · `browserbase-bb--key-metrics-0` Confidence reason: A linked first-party source states the claim. Qualifications: - Reported by: Browserbase - Scope: Automatic feature-request scanning of every closed support ticket and meeting transcript - Denominator: Closed support tickets and meeting transcripts Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) · Preserved content.md, lines 26 ### Key observation 99% of first-response times < 24 hrs Metric · Reported · Medium confidence · `browserbase-bb--key-metrics-1` Confidence reason: The source reports 99% below 24 hours with awkward wording; no sample, timestamps, or method is provided. Qualifications: - Reported by: Browserbase - Scope: Reported support first-response time below 24 hours - Denominator: Support first responses; exact sample and measurement window not provided Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) · Preserved content.md, lines 26 ### Key observation Session investigation: 30–60 min of log-diving → one Slack message Metric · Reported · Medium confidence · `browserbase-bb--key-metrics-2` Confidence reason: A Slack message replaces the manual initiation workflow; the article does not report end-to-end automated investigation latency. Qualifications: - The source does not report the denominator of this figure. - Reported by: Browserbase - Scope: Session investigation initiation: manual log-diving versus a Slack request Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) · Preserved content.md, lines 26 ## Lessons and interpretation ### Lesson One agent with good abstractions beats a fleet of narrow bots Inference · Catalog judgment · Medium confidence · `browserbase-bb--lessons-learned-0` Confidence reason: The catalog derives this observation from the linked sources. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Lesson Separate capabilities from the core loop; domain logic lives in skills + service packages Inference · Catalog judgment · Medium confidence · `browserbase-bb--lessons-learned-1` Confidence reason: The catalog derives this observation from the linked sources. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Lesson Don't trust the model, remove its ability to do wrong: scope tools and services per invocation source Inference · Catalog judgment · Medium confidence · `browserbase-bb--lessons-learned-2` Confidence reason: The catalog derives this observation from the linked sources. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ### Lesson Meet people where they are; Slack is the highest-leverage surface because that's where work already happens Inference · Catalog judgment · Medium confidence · `browserbase-bb--lessons-learned-3` Confidence reason: The catalog derives this observation from the linked sources. Evidence: - Supports · [1] [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) · [Preserved copy](https://github.com/steel-experiments/internal-agents-map/blob/main/archive/sources/browserbase-bb-source-1/content.md) ## Sources 1. [How we build internal agents at Browserbase](https://browserbase.com/blog/internal-agents) - Engineering blog · First party · Evidence - Original URL: - Accessed: 2026-08-12 · Last verified: 2026-08-31 - Preserved copy in the repository: