Changelog
This page provides a concise, up-to-date log of all API changes and improvements since launch.
2026-11-03
Posts data paging: GET /api/iq/keyword_search/{id}/posts_data and GET /api/iq/user_search/{id}/posts_data now accept offset and limit to read a search's posts in pages, and fields=lean to leave out media_url, profile_url and sentiment_reason. Every response carries total, returned, offset and limit, and the keyword endpoint adds platform_counts, the number of posts per platform across the whole search. A malformed page request is answered 400.
Platform filter fixed: the keyword endpoint's platform parameter was documented but not applied, so every platform's posts came back regardless. It now filters as documented, and total counts the filtered posts.
See Posts Data for the parameters and the response fields.
2026-10-22
Search coverage: Keyword and user search responses, the list endpoints and the return_url callback now carry coverage, the status of each requested platform with a reason code when one was left out, and notices, the same in plain language. A search still runs on every platform it can; when none of the requested platforms can be searched, the request is answered 422 with the coverage and nothing is charged. Both fields are additive.
See Keyword Search for the fields and reason codes.
2026-09-08
Default posts per platform is now 100: Keyword and user searches submitted through the API, the MCP servers, and Rolli Agent runs and schedules now collect up to 100 posts per platform when max_post is omitted (previously 50). Pass max_post to override. Existing schedules keep the value they were created with.
max_post accepted on user searches: POST /api/iq/user_search now accepts the max_post parameter.
Default reasoning effort is now max: Rolli Agent runs and schedules created without available_credits now default to 4 (max) instead of 2 (medium).
2026-05-20
Rolli Agent endpoints now available in both MCP servers: All 15 Rolli Agent endpoints (runs and scheduled reports) are exposed as MCP tools — including start_agent_run (which polls until the report is ready and returns the full markdown), list_agent_runs, create_agent_schedule, pause_agent_schedule, and more. Available in @rolli/mcp v1.3.0 on npm and at https://mcp.rolli.ai/mcp.
See Rolli MCP Server documentation for the full tool list and setup instructions.
2026-05-19
Rolli Agent API launched: A new AI-driven intelligence service at https://agent.rolli.ai. Submit a plain-English question and receive a structured markdown report with signals, evidence links, and a confidence rating. Authenticates with the same X-ROLLI-TOKEN and X-ROLLI-USER-ID headers as the rest of the Rolli API.
See Rolli Agent documentation for the full endpoint reference, including runs (create/list/show/cancel/rerun/update/delete) and scheduled reports.
Remote MCP server available: The Rolli MCP server is now also hosted at https://mcp.rolli.ai/mcp with browser-based OAuth sign-in — no API token to manage. The existing @rolli/mcp npm package remains supported.
See Rolli MCP Server documentation for setup instructions and a local-vs-remote comparison.
2026-05-15
Expert search available to producer-permission API users: API keys with the producer permission flag can now use the /api/search endpoint to search for experts.
See Rolli Experts documentation for details.
2026-04-08
Usage endpoint fix: Keyword searches created through the OAuth / Remote MCP flow were not being counted in /api/usage. Counts now correctly include all searches, regardless of how they were initiated. You may see higher numbers than before — these reflect the true usage of your account.
See Usage Data documentation for details.
2026-03-26
Insufficient-credits error code changed: POST /api/iq/keyword_search and POST /api/iq/user_search now return HTTP 402 Payment Required instead of 500 when the account has insufficient IQ credits. Update your error handling to recognize 402 as a credits problem rather than a server error.
2026-01-19
Keyword Searches endpoint expanded: Added max_post parameter to limit number of posts per platform.
See Keyword Searches documentation for details.
Post data endpoint expanded: Emotion and Sentiment data added to Facebook, Instagram, Threads, Bluesky, Weibo, Reddit, and Youtube.
See Posts Data documentation for details on the available data fields for each platform.
2026-01-16
Keyword search and post data endpoints expanded: Now includes support for Facebook, Instagram, and Weibo platforms, enabling broader social media analysis.
See Posts Data documentation for details on the available data fields for each platform.
2026-01-05
Topic tree endpoint updated: The endpoint now returns topic trees from all available platforms except LinkedIn. The response includes a data array containing all topic trees for the search.
Backward compatibility: For now, the first available tree in the data array is also included at the top level of the response object, as in previous versions. This top-level data is deprecated and will be removed on March 1st, 2026. Please update your integrations to use the data array exclusively.