efferent
The accountable data platform

Every system.
One accountable
understanding.

Connect the data you already have, or create a database of your own. Discover what it means. Review how it fits. Keep that understanding alive — and put it to work with control.

Your environment. Our managed databases. Any combination.
Your data stays yours. Your exit stays free.

A living, reviewed mapSynthetic example
▦Warehouse
◈CRM
▤Files
⌁Streams

Meaning, with its evidence.

Types · entities · relationships · definitions

● Settled● Open● Withheld
HarborWorks renewal
$24,000
Source-backed value

Renewal date: UNKNOWN · source dates disagree
Namesake: separate · no approved identity link

Versioned. Recoverable. Exportable.↗ Inspect the evidence
01 / DISCOVER & MAP

Know what your data means.

Inspect schemas, find relationships, review identities and make conflicting definitions visible. Every proposal brings its evidence.

02 / MONITOR & EXPLAIN

Know what changed, and why.

Track drift, staleness, missing values and business conditions against the same reviewed model. Investigate without changing a source.

03 / GOVERN & ACT

Make the next step deliberate.

Preview an operation, check current authority, approve its exact scope and record the result. A finding never grants permission.

Two doors. One platform.

Meet your data
where it lives.

Add intelligence to the systems you already run. Build something new in a managed database. Combine the two on the subset you permit.

↗

Connect existing systems

Your warehouses, databases, tools, files and streams remain authoritative. Run the platform in your own environment, with explicit controls for metadata, values, retention and actions.

Customer environmentRead-only discoveryHybrid by permission
Connect a source →
▦

Create a database

A normal database, with direct and pooled connections. Schedule portable snapshots and approve recovery. Keep intelligence off, or enroll the data you choose. The first managed store speaks PostgreSQL; the platform is built around portable contracts.

Ordinary SQLApplication tablesIntelligence optional
Create a database →
Refusing to guess is the product.

Unknown is a useful answer.
A silent merge is not.

A similar name does not establish identity. A high score does not establish permission. When the evidence cannot settle a question, the platform keeps it open — with a reason you can inspect.

Meaning has a history.

Versioned models and retained inputs let you inspect what was known at a point in time and reproduce what was published.

Numbers show their scope.

Settled, open and withheld coverage accompanies measures. An unresolved value is never quietly converted to zero.

Evidence lives with the data.

Queryable decisions, provenance and value status sit in the database. Ordinary clients can inspect the same foundation.

Breadth is the point.

Start with the work
you need to do.

12 seeded synthetic evaluation estates

Application building, research, finance, operations and beyond. Start without customer data; carry the same reviewed-model workflow into an authorized source.

Across the data world

Connect through the doors
your systems already open.

Protocol families and bounded adapters span relational databases, warehouses, documents, SaaS, object storage and events. The workspace shows installed drivers and the actual result of each authorized connection.

PostgreSQL ecosystemSnowflakeBigQueryDatabricksRedshiftMySQL / MariaDBSQL ServerMongoDBRedis / ValkeyHubSpotStripeSalesforceCSV · JSON · ParquetS3 · GCS · Azure BlobKafka · HTTP events · ordered changes

Adapter implementation, installed driver and live source qualification are separate states. External systems require your credentials and endpoint permission. No customer deployments or universal connector certification are claimed.

Trust includes leaving.

Useful enough to keep.
Open enough to leave.

Export your data, model, decisions and provenance. Restore ordinary SQL without the platform. Keep using your database with intelligence off. There is no punitive export fee.