EN
Back to the archive

The archive · AI & Models · Product decision · 2025–2026

Airweave (YC X25): agent search layer hit 6.6k stars, then repo archived in 2026

YC X25's open-source retrieval layer lets agents search 50+ apps through one API; it drew 164 HN points and 6.6k stars before its repo was archived.

Airweave

The betThat agent builders would adopt one shared, continuously synced retrieval layer over users' apps instead of building per-source connectors.Live

What the business is

Open-source context-retrieval layer for AI agents and RAG systems: connect 50+ apps, tools and databases; Airweave syncs and indexes their data and exposes it through one LLM-friendly search interface (REST API, SDKs, CLI, MCP).

How it started

Airweave was founded in Y Combinator's X25 batch by Lennert Jansen and Rauf Akdemir. On 2025-09-30 they posted 'Launch HN: Airweave (YC X25) – Let agents search any app,' pitching shared retrieval infrastructure: connect apps, databases and documents, and Airweave syncs, indexes and serves grounded, up-to-date context that agents query in a single request instead of chaining fragile tool calls.

What happened

The launch discussion (164 points, 30 comments) pressed the team on permissioning, RBAC and pricing, and on how Airweave differed from enterprise search for humans (Glean) and end-to-end search apps (Onyx); the founders answered that per-user syncs mirroring source ACLs were the near-term answer while richer org-level RBAC mapping was on the roadmap. Public development stayed active — 4,775 commits and 819 forks by mid-2026 — across a cloud app at app.airweave.ai, self-hosting, Python/TypeScript SDKs, a CLI and MCP, with PostgreSQL metadata, Vespa vectors and Temporal workers in the stack.

How it ended up

On 2026-08-29 the owner archived the GitHub repository as read-only, ending public open-source development less than a year after the Launch HN; the material records no successor announcement, and the cloud app page still resolved on 2026-09-05.

Background

Airweave is an open-source context-retrieval layer for AI agents and RAG systems: it connects to apps, tools and databases (50+ integrations), continuously syncs their data, and exposes it through a unified, LLM-friendly search interface that agents query in one request. The company offered a cloud app plus a self-hostable Docker distribution, with SDKs, a CLI and MCP support.

The founding bet was that the growing crowd of agent developers would not rebuild connectors and ingestion pipelines per customer, and would instead adopt shared retrieval infrastructure that handled authentication, syncing, indexing and retrieval once. Founders Lennert Jansen and Rauf Akdemir launched from YC's X25 batch on 2025-09-30 with a 'Launch HN' that positioned Airweave as infrastructure for developers whose agents need grounded, up-to-date context from a user's apps.

The launch drew 164 points and 30 comments on Hacker News, and the repository grew to 6.6k stars, 819 forks and 4,775 commits. But on 2026-08-29 the owner archived the repository as read-only, ending the public open-source project less than a year after launch; the cloud app page still resolved as of 2026-09-05, and the material records no statement about what the company does next.

What has to be true

  • The wedge was real: every agent builder hitting Slack, Drive, Notion and Stripe faced the same ingestion, syncing and indexing work, and Airweave packaged it behind one search endpoint.
  • Per-user syncs that mirrored each source's ACLs gave a credible answer to the launch thread's hardest question — permissioning — even before org-level RBAC existed.
  • Open source plus self-hosting removed the trust objection for developer tools, letting the project accumulate 6.6k stars and 819 forks rapidly.
  • MCP and agent-framework integrations plugged Airweave into the channels agents already used, rather than asking builders to change their stack.
  • The 2026 repo archival shows developer adoption alone did not carry the business: the retrieval layer still needed a paid tier with a moat once platform vendors began adding similar connectors.

What can be applied

Stars measure developer interest, not a business: a 6.6k-star agent-retrieval launch ended with its repo archived in under a year — adoption still needed a paid wedge and a durable moat.

Aftermath

As of 2026-09-05 the airweave-ai/airweave repository remains archived and read-only with 6.6k stars and 819 forks, and the dashboard at app.airweave.ai still resolves; the material records no announcement of a pivot, acquisition or shutdown. What is unambiguous is the end of the open-source project less than a year after its Launch HN.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases