Self-hosted
$0
Recall on polign_db in your own bucket, or on your existing Qdrant. Free for as long as it runs.
- No license file or activation
- No node, vector, or query limits
- Your infrastructure and storage
- Build your own products and services
Recall is free to run on your own infrastructure, on polign_db in your bucket or on a vector store you already have. Paid agreements are a flat annual fee with nothing metered. Ten design partner seats are open at no charge for 60 days.
$0
Recall on polign_db in your own bucket, or on your existing Qdrant. Free for as long as it runs.
$0for 60 days
Then $12k for the next 12 months, once the success criteria are met.
Ten seats. A named Polign engineer wires Recall into your agent with you.
$24kper year
Flat. Billed annually. Nothing metered.
Polign on your team, for one number you can budget.
Custom
Ship polign_db inside the product you sell.
Each release is free to run under its included license. For self-hosting, infrastructure costs are paid to your provider separately. Paid agreements are annual only; there is no monthly plan and no usage line on the invoice. Design partner seats start with a short application. We agree one use case, written success criteria, and a named owner on your side, and both sides sign an order form before day one. The seat runs for 60 days. If the criteria are met, we invoice $12k for the following 12 months under those terms; if not, it ends with nothing owed. Production Agreements are signed the same way and invoiced once a year. The agreement covers how Recall behaves on any backend we support. With polign_db as the store, storage-engine support, upgrade help and the recovery runbook are included; a Postgres or Qdrant you run stays yours to operate. Building a service on polign_db that you run yourself is covered by the free license. Handing the software itself to your customers needs the OEM agreement.
On polign_db the infrastructure bill is yours, and it is small: these numbers were measured with polign_db's benchmark harness against same-region Amazon S3. Full methodology on the benchmarks page. For a live example, read the case study: one Wikipedia index in S3, from an $11 box to 860 searches a second.
One pip install to try it. Your own bucket when you share it.