generated: '2026-08-01'
method: probed
source: https://mcp.datarails.com/mcp (response headers) + Datarails FinanceOS MCP documentation
description: >-
  Datarails signals rate limiting on the FinanceOS MCP server with x-ratelimit
  headers, and separately documents hard result-size ceilings per tool family.
  The two are different things and both are recorded. No rate limits are
  documented for the Data Gateway Service upload endpoint.

headers:
  - name: x-ratelimit-limit
    observed_value: '120'
    surface: https://mcp.datarails.com/mcp
  - name: x-ratelimit-remaining
    observed_value: '119'
    surface: https://mcp.datarails.com/mcp
  - name: x-ratelimit-reset
    present: false
    note: No reset header was returned, so the window length is not advertised.
retry_after: false
retry_after_note: >-
  No Retry-After header on rate-limit responses. A retry_after_seconds field
  does appear inside async poll results, but that is job scheduling, not
  throttling.

rate_limits:
  - name: FinanceOS MCP request rate
    limit_count: 120
    period: unknown
    unit: requests
    scope: per client, inferred from the decrementing x-ratelimit-remaining header
    surface: https://mcp.datarails.com/mcp
    method: probed
    note: The window is not published; only the ceiling is observable.

result_ceilings:
  - name: Filtered row fetch
    tools: [get_data_by_alias, get_data_by_id]
    limit_count: 500
    unit: rows per page
    source: https://support.datarails.com/hc/en-us/articles/25849710214556-Datarails-FinanceOS-MCP-Server-Technical-Documentation
  - name: Custom query
    limit_count: 1000
    unit: rows
    source: https://support.datarails.com/hc/en-us/articles/25849710214556-Datarails-FinanceOS-MCP-Server-Technical-Documentation
    note: 'Documented as "Custom queries: up to 1,000 rows".'
  - name: Aggregations
    limit_count: null
    unit: rows
    source: https://support.datarails.com/hc/en-us/articles/25849710214556-Datarails-FinanceOS-MCP-Server-Technical-Documentation
    note: Documented as having no row limit; this is why Datarails tells agents to prefer aggregation over pagination.
  - name: Serialized response size
    limit_count: 100
    unit: KB (approximate)
    source: https://github.com/Datarails/dr-claude-code-plugins-re/blob/main/CLAUDE.md
    behaviour: >-
      Exceeding it returns HTTP 200 with a truncation envelope
      {data, truncated, total_rows, returned_rows, guidance} where data is an
      incomplete prefix.

token_limits:
  - name: JWT access token lifetime
    value: 5 minutes
    behaviour: Auto-refreshed for aggregation; manual refresh required every 20K rows when paginating.
    source: https://github.com/Datarails/dr-claude-code-plugins-re/blob/main/CLAUDE.md

x-evidence:
  fetched: '2026-08-01'
  url: https://mcp.datarails.com/mcp
  http_status: 401
  headers_seen: [x-ratelimit-limit, x-ratelimit-remaining, x-request-id, x-trace-id, strict-transport-security]