BetaWeekly digests are a beta — we're trying something new. Feedback welcome.
Agents get richer tools while browser sessions get organized
November 9–15, 2026
Firecrawl pushed its agent surface hard this week, expanding its company and AI/ML research data into ChatGPT and Claude and shipping a rebuilt MCP with thousands of tools, while Browserbase added session labels and dashboard-managed secrets for keeping automation runs tidy and secure.
The week at a glance
Replay the weekBiggest releases
- 01Firecrawl MCP v3 adds 3000+ tools with provider waterfalls and 60x rate limits
Firecrawl - 02Firecrawl brings an AI/ML research index to ChatGPT and Claude
Firecrawl - 03Firecrawl job search tools match resumes to 7M+ open roles in ChatGPT and Claude
Firecrawl - 04Browserbase adds session labels for organizing and filtering runs
Browserbase - 05Browserbase lets you create Secrets from the dashboard
Browserbase
Firecrawl turns itself into an agent research layer
Firecrawl spent the week widening what an agent can actually look up. The biggest structural change is a rebuilt MCP that gives agents over 3,000 tools with automatic provider waterfalls, alongside dramatically higher rate limits, more included usage, and parallel subtasks — and it works across every harness rather than locking you into one client. That matters because the data products layered on top are now reachable from wherever your agent already lives.
Chief among those is a much larger company graph: Firecrawl Alexandria now spans more than 75 million companies, returning headcount, revenue, funding, and tech stack in roughly half a second. For anyone building prospecting or enrichment flows, that turns what used to be a scraping-and-parsing pipeline into a single fast lookup.
Two more indexes target specific agent workflows. A new AI/ML research index lets ChatGPT, Claude, or any harness find and read full papers, including ones posted the same day — a real boon for keeping up with fast-moving research. And job search tools match a resume against 7M+ open roles while pulling executives to contact from a 250M+ person database, complete with context on what they care about before outreach. Together these read as a deliberate bet: don't just let agents browse, give them curated indexes that answer questions directly.
Browserbase makes long-running automation easier to manage
Browserbase's two updates are both about the operational reality of running lots of browser sessions. Session labels let you tag sessions by workflow, environment, website, or automation version, then filter on them in the dashboard — and they can be set via the API or the UI, shared across a project, and renamed or removed. If you've ever stared at a wall of anonymous sessions trying to figure out which one came from the staging run, this is the fix.
The other addition, creating Secrets natively from the dashboard, stores encrypted credentials you can use at runtime in Functions to call external APIs. That removes a chunk of credential plumbing from the setup path — no more routing secrets through config by hand.
Where this leaves the collection
The through-line this week is agents doing real work with less glue code. Firecrawl is increasingly a research backend you point an agent at rather than a scraper you orchestrate, and Browserbase is filling in the housekeeping — labels and secrets — that separates a demo from a fleet of automations you can actually operate. Both products are quietly optimizing for scale rather than first-run experience.
Releases covered6
- Firecrawl MCP v3 adds 3000+ tools with provider waterfalls and 60x rate limits (opens in new tab)
- Firecrawl Alexandria covers 75M+ companies with firmographic data (opens in new tab)
- Firecrawl brings an AI/ML research index to ChatGPT and Claude (opens in new tab)
- Firecrawl job search tools match resumes to 7M+ open roles in ChatGPT and Claude (opens in new tab)