Keido API

Your systems, extended.

You already have platforms, portals and pipelines that cannot be replaced. The Keido API puts search, mapping, evaluation and data ingestion inside them, so your teams get Keido’s capability in the tools they already open every morning. It runs within your systems, in your environment.

Book a technical scoping call

Request the API documentation

For technical teams

Six endpoints. Every call logged.

REST over HTTPS, key and token authentication, deployable in your cloud, hybrid or on-site. Batch and streaming modes where they make sense.

Endpoint	What it does	Typical use
Search	Hybrid semantic and keyword retrieval with context expansion and full citation trails	A grounded assistant over a research, scientific or government corpus
Embedding	Text to vector, batch or streaming, versioned, compatible with common vector stores	Feeding your own clustering, classification or ML pipeline
Evaluation	Score outputs against reference answers or custom rubrics: precision, recall, semantic similarity, bias and safety checks	Gating a model change in CI, or scoring a summariser in production
Visualisation	UMAP 2D and 3D projections with clustering, taxonomy overlays and tenant-aware access	Embedding a portfolio map in an internal portal or dashboard
Data Pipeline	Ingestion, incremental update, metadata cleaning, versioning and secure transfer with full logging	Keeping an index current from a records system or data warehouse
Governance layer	Access control, observability and audit across all endpoints, aligned to your framework	Meeting internal review and external compliance requirements

A first integration

One Search call, with its citation trail.

# Request
curl -X POST https://<your-deployment>/search \
  -H "Authorization: Bearer $TOKEN" \
  -d '{
    "query": "coastal erosion monitoring since 2018",
    "top_k": 3,
    "expand_context": true
  }'

# Response (truncated)
{
  "results": [{
    "id": "proj-20814",
    "title": "Remote sensing of shoreline change",
    "score": 0.91,
    "citations": [{
      "source": "grants/2019/GR-4471.pdf",
      "page": 6,
      "passage": "…annual LiDAR surveys across 14 sites…"
    }]
  }],
  "request_id": "req_7f3a…"
}

Deployment options

Three ways to run it.

Option	Where it runs	Who manages it
Your cloud	Your AWS, Azure or GCP account, in region	Keido deploys, you or Keido operate
Hybrid	Compute in your cloud, sensitive data on-site	Shared
On-site	Your hardware, air-gapped if required	You operate, Keido supports

How the endpoints fit together

Adopt one endpoint, the set, or run agents on top.

Pipeline feeds Embedding, and Search and Visualisation read from it. Enrichment and scan loops keep the index current.

Agents run a four-step loop, retrieve, reason, act and check, calling Search, Embedding and Pipeline until the check passes. Every run happens inside an evaluation harness, where the Evaluation Engine scores it against your rubrics, guardrails and test cases. Governance wraps all of it, agents included.

Getting access

From call to deployment.

Scoping A 30-minute technical scoping call with your engineering lead
Sandbox Sandbox credentials against a sample dataset, within a week
Deployment Into your environment, typically two to four weeks depending on data sources

Next step

Talk to an engineer about your first integration.

Book a technical scoping call

Request the API documentation