Introducing Firecrawl Developer Index, an index for supercharging coding agents. Search 70M+ primary sources including repos, docs, & issues with the highest recall of any coding-specific index. Ensure agents ship correct, up-to-date code every time!

Why a coding-specific index? Your agents can answer their own questions about code behavior, API contracts, error messages, & known bugs from the correct primary sources. That means issues, merged PRs, & READMEs from the most popular public repos & docs sites.

Our Developer Index hits 63% recall@10 across 1,179 real developer queries spanning repos, docs, and PRs. That beats the next best external provider by ~10%.

Ready to try it out? For best performance we recommend using the Firecrawl CLI or MCP with the companion skill, which you can install with: npx -y firecrawl-cli@latest setup developer-index

Also you can run the benchmark yourself from the repo! Let us know if you'd like to be included! Note: Our benchmark ships with half the queries, and we hold out the rest so nobody can train on the test.

I've been messing around with this internally for a bit. Glad everyone else can see how awesome it is.

this is the 3rd specialized index they've shipped this cycle, research index for arxiv, now this. reads less like a random feature drop and more like firecrawl building a set of domain-specific verticals on top of the general crawler, which is the smarter move than one giant undifferentiated index

Wish my agents could make videos like that

My agents are going to love this!

How does it compare to context 7?

All curation grows until it requires search, and all search grows until it requires curation

This is very useful. Agents are going to love this. Way better than guessing from stale docs page 🔥

coding agents didn’t need another wrapper, they needed fresh primary sources and a benchmark that proves retrieval actually works. 63% recall@10 got our attention

Wanted to build this for sooooo long Thanks!

Firecrawl is on its way to do everything and everything so cool

Ouuuu

Recall@10 is useful, but daily refresh creates a moving target. Does DevDex pin each provider's index state and query timestamp so a rerun separates ranking quality from corpus freshness?

70M sources is actually huge

Amazing web

We need b2b buying intent tracking @firecrawl

Better retrieval makes applicability more consequential. A coding agent can surface a README, merged PR, open issue, and versioned docs for the same behavior. Every source can be authentic and still describe a different version, state, or scope. The harness still has to bind that evidence to the task: what applies to this repo state, what has been superseded, and how conflicting sources should affect the change. Recall gets the right evidence into view. Correctness still depends on deciding what actually applies.

Can't stop. Won't stop.

highest recall measured against what. is there a public eval for this or internal numbers only

i'm sure it can find 70m sources and still pick the first outdated readme.

Recall@10 is useful, but daily refresh creates a moving target. Does DevDex pin each provider's index state and query timestamp so a rerun separates ranking quality from corpus freshness?

This removes a major failure mode: coding against stale internal knowledge. But retrieval improves the evidence available to an agent; it does not prove the resulting code is correct. A production run should bind the repository commit, documentation versions, retrieved sources, generated patch and test results into one verifiable record. Fresh sources help an agent write better code. Execution receipts prove what it actually changed and whether the declared checks passed.

70 million sources indexed, huge

how much of the 70m is issues versus repo code

Ouuu

7OM+ sources with highest recall? That's impressive, but can it really keep agents from shipping outdated code?

high recall helps, but version-aware ranking matters more when deprecated docs outrank current apis

cool vid!

Love those art direction

Recall is the easy half. The hard half is that half of those 70M sources contradict each other and the index has no idea which repo's README is three breaking changes stale.

It would be so good for the tool to include reddit, x, discord, stack overflow, etc.