Every source, one governed line to your agents.

Aggre runs each source through one pipeline, from fetch to publish, and serves versioned datasets over MCP and REST. Your key only rides the lines you choose.

Free with an Orbbit account, one dataset included.

Line map: Legal, Healthcare, Markets, Regulatory news, Your sources each run through Aggre's pipeline and end at two termini, MCP and REST.

The pipeline.

The catalog's lines and your own sources run the same pipeline, in the same order, under the same governance.

Fetch

A connector pulls the source: HTTP with pagination, headers and request bodies, objects from S3 or R2, or a custom worker that claims the job and posts records. Preview a source before its first run.

kind: http · s3 · r2 · custom

Normalize

Records are read out of the payload at their records path and given one shape. The schema is inferred from live samples and versioned with the dataset.

records_path · external_id_path

Dedup

Each record is keyed by its external id, or by a hash of its content when the source has none, so fetching a source again updates what changed instead of duplicating it.

key: external id, or content hash

Validate

Each run counts what it accepted and what it rejected, and the error log keeps the reason for every rejection.

accepted · rejected · error log

Publish

The run publishes a new dataset version with a quality report. Sensitive fields named in the dataset's governance are masked on every way out.

dataset version · quality report · masking

Two termini, the same data.

Every published dataset is served over MCP for agents and over REST for everything else, behind the same key and the same scopes.

MCP

aggre-mcp · stdio

For agents. Point any MCP client at the aggre-mcp server; it finds every tool with tools/list.

15 tools · .mcp.jsonHover to open

Tools

list_datasets
search_dataset
get_record
get_dataset_schema
export_sample
export_dataset
retrieve
retrieve_batch
search_judgments
get_judgment
lookup_judgment
get_politician_trades
get_investor_holdings
get_insider_trades
who_traded

.mcp.json

{
  "mcpServers": {
    "aggre": {
      "command": "/path/to/aggre-mcp",
      "env": {
        "AGGRE_API_BASE": "https://aggre.orbbit.ai",
        "AGGRE_API_KEY": "aggre_sk_…"
      }
    }
  }
}

REST

/v1 · x-aggre-api-key

For everything else. The same tools and records over HTTPS, with the key in x-aggre-api-key.

5 endpoints · curlHover to open

Endpoints

POST /v1/mcp/callevery MCP tool over plain HTTP
POST /v1/searchsearch across the datasets a key can read
POST /v1/exportexport records as JSON or CSV
GET /v1/datasets/{id}/recordspage through one dataset
GET /v1/records/{id}one record, masked

Legal line, with curl

curl -X POST \
  https://aggre.orbbit.ai/v1/mcp/call \
  -H "x-aggre-api-key: aggre_sk_…" \
  -H "content-type: application/json" \
  -d '{ "name": "retrieve",
      "arguments": {
        "query": "公務員收賄罪的量刑",
        "corpus": "both",
        "jurisdiction": "TW",
        "top_k": 5 } }'

Pick a plan, then pick your lines.

A key reads only the datasets you give it, up to your plan's allowance. Calls to the data tools count toward one monthly total for the whole organization, across all of its keys.

Requests per month: 10,000
Datasets for your organization: 1 of 1

Included with your Orbbit account.

Datasets for this key

Legal

Healthcare

Markets

Regulatory news

The key reads only what you pick; change it any time.

curl -X POST \
  https://aggre.orbbit.ai/v1/mcp/call \
  -H "x-aggre-api-key: $AGGRE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "name": "search_dataset",
    "arguments": {
      "dataset_id": "taiwan-judgments",
      "query": "公務員收賄罪的量刑",
      "limit": 10
    }
  }'

The key goes in x-aggre-api-key, or in an Authorization: Bearer header.