eiye.ai · semantic surface · source-available (BSL 1.1)
Organizations want AI agents on existing data without copying a lake, standing up a warehouse, or leaking people. eiye is a pass-through governed query layer: register a source, discover schema, answer MCP or REST queries with redaction, access control, and audit on the way through.
01
eiye sits in front of data you already run. It is a semantic surface, not another place to put the data.
list_datasources, get_schema, query_datasource — or REST02
Nothing is extracted into eiye for safekeeping. The query walks this path and returns. The source remains the source of truth.
Path: register → discover → query → redact / ACL / audit → optional cache → source.
03
eiye does not require an LLM. Agents call tools. A natural-language conductor is planned; it is not in the product today.
list_datasources, get_schema, query_datasourceCommercial-license features: SSO, compliance reports, multi-tenant control plane, advanced analytics.
04
Free to run up to 5 sources and 1,000 queries/month — that limit is the license grant itself, not a trial. Paid tiers raise it. Subscribe on Stripe. Enterprise is custom.
| Tier | Datasources | Queries / mo | Price | Buy |
|---|---|---|---|---|
| Free | 5 | 1,000 | $0 |
GitHub |
| Starter | 15 | 25,000 | $100/mo |
Subscribe |
| Pro | 50 | 250,000 | $500/mo |
Subscribe |
| Business | 150 | 1,000,000 | $2,000/mo |
Subscribe |
| Enterprise | Custom limits, SSO, compliance reports, multi-tenant control plane, advanced analytics | Custom | Contact | |
The Free tier is a license grant, not just a limit: the BSL 1.1 permits production use up to 5 datasources and 1,000 queries/month at no charge. Beyond that, or to offer eiye as a hosted service, requires a commercial license. Starter, Pro, and Business subscribe on Stripe; we email the license after payment. Paid tiers also unlock SSO, compliance reports, multi-tenant control plane, and advanced analytics. Enterprise stays a conversation: ask about a commercial license.
05
Primary path is the repo. One script registers a source, discovers schema, runs a governed query, and prints the MCP connect command.
git clone https://github.com/eiye-ai/eiye-db.git
cd eiye-db
scripts/quickstart.sh --name mydata --type filesystem --root /path/to/a/folder
# or: --type postgresql --dsn postgresql://user:pass@host/db
# or: --type rest_api --url https://api.example.comConnect an agent (from the README):
claude mcp add eiye-db -- /path/to/backend/.venv/bin/python -m eiye_db.mcp_serverTools the agent gets: list_datasources, get_schema, query_datasource — read-only, PII-redacted, audited.
06
No. eiye is pass-through. Data stays in your sources. There is no warehouse, no ETL job, and no copy of the lake. You self-host it. Queries go to the source; redaction, access control, and audit sit on that path.
No. Non-goals: ETL, warehouse, BI dashboards, email marketing. eiye is a governed query layer in front of systems you already operate.
No. eiye does not require an LLM. Agents call MCP or REST. A natural-language conductor is on the roadmap and is labeled as such — it is not required to run the surface today.
The full source is public and auditable — connectors, PII engine, access control, audit, MCP, REST, React console — but it is not open source in the OSI sense, and we do not call it that. eiye is licensed under the Business Source License 1.1: read it, run it, modify it, and use it in production free of charge up to 5 datasources and 1,000 queries per month. Larger production use, or offering eiye to third parties as a hosted service, requires a commercial license. Each release converts to Apache-2.0 four years after publication, so nothing is enclosed permanently.
On the query path. Baseline redaction is regex. Optional spaCy NER for names and locations. Combined with access control and a recorded audit trail.
PostgreSQL, a filesystem root (CSV, text, PDF, XLSX), and REST GET. That is what is built now — not a promise list.