Tessera DB
PX-01
By problem
Built for data where the connections, the history and the proof all matter.
Four kinds of work where running several databases hurts most: following money, grounding AI answers, proving what the data said, and seeing one customer across many systems. Each page says what the work needs, what Tessera DB does for it, and what it does not do yet.
01
03
Pick your problem.
02
03
What buyers ask first.
You are new. Why should we bet on you?
We publish what is built, what we have measured and what is not built yet, and you can read it before talking to us. The capability table in the kit is generated from the server's own register, and every partial feature names its gap. The Free tier is a free licence, not a paid trial.
Do you have customers we can talk to?
Not publicly. We will not invent references. What we can offer is a scoped evaluation and direct access to the engineers who built it. Write to contact@par2labs.com.
Is Tessera DB open source?
No. The Free tier is a free binary with a free licence key, and the source is available on request. Pro, Enterprise and Enterprise+ unlock on the same binary with a signed key.
How do I connect from my language?
Over plain HTTP and JSON from any language; the kit has curl, Python, JavaScript, Java and Go examples. The Postgres wire protocol works with psql, psycopg and JDBC for reads, and Bolt works with the official Neo4j drivers for Python, JavaScript, Go and Java. No client package is published to npm or PyPI yet.
Does it replace PostgreSQL?
Not fully. The SQL is a large subset, and the kit's query guide lists what is missing — for example ROUND and CREATE INDEX — and one SQL statement runs per request. If you need full PostgreSQL compatibility, keep PostgreSQL for that part.
Where does it run?
On your own hardware, as one server binary. On Linux it needs glibc 2.34 or newer and OpenSSL 3 (Debian 12, Ubuntu 22.04, RHEL 9 or newer). A query needs no external service: the embedder is built in, and you can point it at your own model.
03
03
Two promises, whatever the problem.
01
One record, not seven copies
The graph, the rows, the documents, the search indexes, the history and the audit trail read the same record. There is no sync job to fall behind, because there is nothing to sync.
02
The limits are written down
Every page says what is not built yet. A database that says only what it does is telling you less than it appears to.