Synth Index: free public Search on the web

Current v0.2 availability: Index ships through the API, Python SDK/CLI and MCP. The browser Search described in this September 28 release record has been retired; browser Index product UI is deferred to v0.3. The website provides account setup, API keys, payments and basic telemetry. Public API FAST and DEEP Search are free only when the deployed capability is enabled. v0.2 production qualification remains a separate release gate; this historical entry does not announce its completion. Use the Index quickstart for the supported client surface.

This release follows Synth Index alpha v0.1. It adds free public Search in the browser. Signed-in API, SDK and MCP Search keep working. With synth-ai>=0.21.1, the CLI, SDK and Index MCP server also support anonymous public Search. Authenticated Search uses the paid route; choose the public route explicitly to search for free without an API key.

Index collects grounded mētis for research engineering: practical findings about what worked, what failed, and the conditions under which a result holds. Search returns cited research evidence and a summary; inspect the source and its limitations before applying a finding to your own work. Deep remains beta and may return no matching evidence even for an answerable question.

What is new

Public Search on the web. Open Synth Index and search the public Index. You do not need an account or an API key.
Fast and Deep modes. Fast returns results quickly. Deep takes longer and digs further. Each result cites the exact published revision it came from.
Free while public Search is on. The page shows the price for each mode before you search, taken from the live service. It shows public Search as switched off when the service turns it off.
Published retention. The page shows the retention terms currently in force. Public query and response content is not kept in durable storage. Encrypted processing state expires within 60 minutes. Operational and billing metadata that does not include your query is kept.
Limits you can see. Public Search has per-visitor limits per minute and per day, shared limits, a shared daily budget and a cap on concurrent Deep searches. When you reach a limit, the page says which one and counts down the Retry-After time before you can search again.
Deep searches you can resume or cancel. A Deep search keeps running on the server while the page waits for it. If the connection drops or the page stops waiting, you can resume the same search instead of starting a new one. Cancel stops the search on the server. Leaving or reloading the page loses this browser's access to the result, but it does not cancel the search.

Not in this release

An endpoint that writes a direct answer to your question.
Contributing research, reviewing it, publishing it, public profiles or contributor credits. None of these is open in this release.
Contests.
Searching private collections from the browser.

Get started

For current API, SDK/CLI and MCP use, follow the Index quickstart.

What is new

Public Search on the web. Open Synth Index and search the public Index. You do not need an account or an API key.

Fast and Deep modes. Fast returns results quickly. Deep takes longer and digs further. Each result cites the exact published revision it came from.

Free while public Search is on. The page shows the price for each mode before you search, taken from the live service. It shows public Search as switched off when the service turns it off.

Published retention. The page shows the retention terms currently in force. Public query and response content is not kept in durable storage. Encrypted processing state expires within 60 minutes. Operational and billing metadata that does not include your query is kept.

Limits you can see. Public Search has per-visitor limits per minute and per day, shared limits, a shared daily budget and a cap on concurrent Deep searches. When you reach a limit, the page says which one and counts down the Retry-After time before you can search again.

Deep searches you can resume or cancel. A Deep search keeps running on the server while the page waits for it. If the connection drops or the page stops waiting, you can resume the same search instead of starting a new one. Cancel stops the search on the server. Leaving or reloading the page loses this browser's access to the result, but it does not cancel the search.