Why we built Headband

This post is from May 2026. For what Headband is today, start at the home page.

Most hosted search runs one engine.

Algolia assumes you want their proprietary engine. Typesense assumes you want Typesense. Elasticsearch assumes you want Elasticsearch. And if you're building anything non-trivial, that assumption breaks down fast.

Your product catalog needs typo-tolerant search, and Meilisearch is great at that. Your docs and RAG pipeline need real vector search, and Elasticsearch has it. Your autocomplete needs consistent latency on every keystroke, and Typesense is built for it.

But none of these tools let you mix. You end up with three APIs, three SDKs, three dashboards, three billing relationships, and three different ways to manage API keys.

What Headband actually is

Headband puts several search engines behind one account. One API, one SDK, one dashboard, but you pick the engine per index. Each engine's own API is also served at api.headband.dev, so you can keep the client you already use.

Create a products index on Meilisearch. Create an autocomplete index on Typesense. Create a docs index on Elasticsearch. Same API key. Same client code. Same filter syntax. The engines still differ in what they support, and the engine compatibility page says how.

And if you already run your own Meilisearch, Typesense or Elasticsearch cluster, Headband doesn't replace it. Bring your own engine, and the search index stays in your infrastructure. Headband keeps only what it needs to route and run searches: the engine's address and credentials, encrypted, plus hashed keys, usage and analytics.

Why now

Two things changed that make this approach viable:

First, Meilisearch and Typesense matured. Now there are genuinely good alternatives, each with different strengths. The market fragmented, but the tooling didn't keep up.

Second, AI agents need retrieval. Every agent framework needs a way to search your data: RAG pipelines, MCP servers, llms.txt. Headband ships all of these out of the box, across any engine. Your agent doesn't need to know whether the index it's searching runs on Meilisearch or Elasticsearch.

What's live today

• Per-index engine routing (Meilisearch, Typesense, Elasticsearch and Visor)
• Bring your own engine: point a Meilisearch, Typesense or Elasticsearch index at your own cluster
• Unified search API with engine-agnostic filter syntax
• React SDK (@headband/react) with InstantSearch-inspired components (familiar, but not a drop-in replacement)
• A read-only MCP server for AI agents
• RAG-ready search endpoints
• llms.txt generation
• Scoped API keys (admin + search, per-project)
• Search analytics in the dashboard

What's next

Update, October 2026: A/B testing has since shipped as hosted experiments (/v1/experiments). See the A/B testing docs. If you're interested, try it free at headband.dev. You can also point any index at your own Meilisearch, Typesense or Elasticsearch cluster, so the data stays in your infrastructure.