Report rows, writes, CSV and the rest of the screens
Reports. Three report routes: brands, domains and URLs, broken down by any of date, engine, segment and prompt, every figure with its band and sample size. Budgets are counted in prompt-engine-days; with
date as a dimension a report pages through up to a year. See Reports.Writes. Add, re-word and archive prompts; track and stop tracking competitors by brand id; add and remove aliases; set claimed segments and engines. Two new scopes, prompts:write and projects:write, which Workspace settings offers unticked. A write acts as the person who made the key. See Writes.CSV output. format=csv on the projects, prompts, answers and engine searches lists and on the three reports, with each figure’s band and sample size in their own columns. See CSV output.More reads. The engine catalogue, a project’s configuration (brand, aliases, competitors, engines, segments), engine searches, Objections and one objection’s sources, and Crawlability.MCP. Nine more read tools to match: get_project, list_engines, get_brand_report, get_domain_report, get_url_report, list_searches, list_objections, get_objection and get_crawlability. The MCP server still has no write tools.Also. The API’s own address, https://api.citino.ai/v1, and a published OpenAPI 3.1 description of every route, which the API reference is built from. A request that runs out of time is now 503 query_timeout rather than 500. A key whose workspace is suspended or cancelled is now refused with 401.Crawler activity
Crawler activity. AI crawler visits to your own site, from the server logs you send: visits by crawler and by page, with verified and claimed visits kept apart, beside how often each page is retrieved and cited. Needs
sources:read.MCP. get_crawler_activity, its tool.Every screen's reports, and signing in from Claude and ChatGPT
Seventeen reads, one per screen: the daily trend, visibility by engine and by segment, what moved, the prompts you are losing, Prompt detail, Sources and Source detail, Attribution, Alerts, Experiments and one experiment, Gaps, Claims, an agency’s monthly reports and one month, and the metric definitions the app’s
? shows.Three new read scopes for them: sources:read, attribution:read and alerts:read. The other new routes use metrics:read, so an existing key with that scope now also reads work-item titles and reasons, claims, experiments and an agency’s monthly reports.MCP. A tool to match each new route, from get_trend to explain_metric. A tool that cannot answer (a missing argument, an unknown id) now says why in its result, so the AI can correct itself. And Claude, ChatGPT and Claude Code can now connect by signing in, with no key to copy. See Connect Claude or ChatGPT.The first version of the API
The first routes: your projects, your headline figures (visibility, share of voice, position and sentiment, each with its band and sample size), the league table, your prompts, the answers collected, and one answer with its text.The MCP server, with five read tools:
list_projects, get_metrics, get_league_table, list_prompts and list_runs. No tool returns an engine’s answer text.