# Flarea API v1 AI-first live show platform. This API is designed to be driven by agents (Claude Code, Codex, cursor agents, plain curl). Everything a host can do, your agent can do. Base URL: https://api.flarea.ai/api/v1 Auth: Authorization: Bearer (get it from your dashboard → API) ## Endpoints GET /api/v1 this document GET /api/v1/events list your events (+registration counts) POST /api/v1/events create event (AI also writes the signup-page copy) body: { "title": "...", "brief": "guests, topics, key points", "starts_at": "2026-07-06T18:00:00Z", // optional ISO "theme": "editorial|collage|retro|midnight", // optional "social": "@yourhandle", // optional "generate_show_plan": true, // default true: AI builds segments, ticker, questions + signup copy "guests": [ { "name": "...", "title": "...", "company": "...", "intro": "host reads on air", "emcee": "MC walk-on line", "photo": "https://…" } ], "emcee": { "voice": "onyx|echo|fable|nova|shimmer|alloy", "intro": "show-opener script" }, "reminders": { "h24": true, "h1": true, "live": true, "changed": true }, "signup": { "headline": "...", "sub": "...", "config": { "cta_text": "...", "show_agenda": true } }, "brand": { "accent": "#FF4D00", "logo": "https://…" } } → returns slug + links (studio, signup page, legacy stage-request door, viewer link) GET /api/v1/events/{slug} { event: {…config, reminders, emcee_voice, guest_count, registrations}, sessions: [...], cta_clicks: [...] } PATCH /api/v1/events/{slug} update any of: { title?, description?, starts_at?, theme?, social?, signup?, emcee?, reminders?, guests?, brand?, plan?, questions? } plus the setup fields listed in the OpenAPI document (timezone, ends_at, host, language, after, watch, survey, …). Dates are ISO 8601 WITH a zone; moving starts_at moves ends_at with it. guests is MERGED onto the roster (send only the guests to add or change; { "id": "…", "remove": true } removes one) and needs the guests:write scope. DELETE /api/v1/events/{slug} delete event GET /api/v1/events/{slug}/registrations attendee list (name, email, created_at) POST /api/v1/events/{slug}/registrations add an attendee body: { "name": "...", "email": "..." } POST /api/v1/events/{slug}/registrations/bulk setup-page equivalent for people already signed up elsewhere: { "records": [{ "name": "...", "email": "...", "marketing_consent": false }], "enrollment_mode": "external_signup", "send_confirmation": true, "source_attestation": { "source_type": "opt_in_list", "source_detail": "LinkedIn Lead Gen form ..." } } max 5 per call; requires registrations:write + communications:send; dedupes, honors unsubscribes, sends confirmation + calendar invite, and enrolls reminders. DELETE /api/v1/events/{slug}/registrations remove an attendee ?email=person@co.com GET /api/v1/events/{slug}/recordings program file, per-speaker tracks, rendered clips (public URLs) GET /api/v1/events/{slug}/analytics deep post-show analytics — the same numbers as the in-app report: registered, attended, peak_concurrent, duration_min, avg_watch_min, completion_pct, cta_clicks_total, chat totals (real vs synthetic), replay_views, nps {avg,count}, polls, retention (per-minute curve, chartable as-is), attendees (per-person watch time + engagement, sorted by minutes — join on email for lead scoring) GET /api/v1/events/{slug}/studio live status (host present, people) + the full command catalog POST /api/v1/events/{slug}/studio drive the LIVE show (requires the host on air; 409 otherwise): { "action": "set_layout", "layout": "duo|present|solo|grid" } { "action": "go_scene", "index": 0 } // -1 = last scene { "action": "add_scene", "scene_type": "hot_take", "layout"?: "duo" } { "action": "confetti" } // one-shots: show_deck hide_deck next_slide // prev_slide quote_card confetti applause flowers show_whiteboard // show_video clear_stage zoom_in zoom_out zoom_reset immersive_on // immersive_off music_start music_stop { "command": "bring up the slides and hit confetti" } // natural language, // the same voice-agent path as saying "Flarea, …" on air { "scene": { ... } } // raw scene-state patch (advanced) ## Studio control (live-show API) The studio endpoint gives an agent FULL control of a show while it's on air — layouts, run-of-show navigation, slides, sound, effects — through the same execution paths as the host's own voice commands. Commands are validated server-side and injected into the live room; the host's studio applies them instantly. A show with no host on air returns 409 not_live. ## Shorts & clips (post-show API — full control of the native clip engine) GET /api/v1/events/{slug}/shorts everything: analysis (ranked moments + predicted virality scores), clip slots used/cap, live render jobs (stage + progress), finished clips (MP4 + thumbnail URLs), and a self-describing catalog of every template/style knob POST /api/v1/events/{slug}/shorts/analyze rank the episode's best moments (cached per source; uploads with no transcript whisper on the worker → {pending:true}, poll the GET). Optional { "auto_render": true, "n": 3 } also renders the top N picks in one call. POST /api/v1/events/{slug}/shorts/render render or RE-render (edit) one clip: { "start_sec": 120, "end_sec": 150, // 5-120s "template_id": "editorial_motion|speaker_focus", "orientation": "vertical|horizontal|square", "caption": { "preset": "impact|pill|block|underline|bracket|minimal", "size": "sm|md|lg" }, "motion": "calm|standard|energetic", "highlight_hex": "#FF4D00", "logo": true, "kicker": "PIPELINE MATH", "title": "…", "speaker": "…", "removeSilences": true, "improveSound": true, "removeFillers": false, "focusMode": "auto|pin|off", "focusSpeaker": "Dana", "focusTile": 0, "followCta": false, "followHandle": "@you", "timerBadge": false } Re-rendering the same moment with new style never consumes another clip slot. focusMode "pin" holds one speaker for the whole clip (by name, or by tile index when name tags are unreadable); an unmatchable pin renders the full frame — never a guessed face. POST /api/v1/events/{slug}/shorts/prompt { "prompt": "clip of the right speaker summarising the interview, fast cuts, no filler words" } One request → parsed plan → brief-ranked moment → one queued render. Response: job_id, the resolved plan/segment, and notes for anything that degraded. POST /api/v1/events/{slug}/shorts/jobs { "job_id": N, "action": "cancel" } (only while queued) POST /api/v1/events/{slug}/shorts/segments { "segment_id": N, "action": "archive" } free the slot { "segment_id": N, "action": "note", "note": "…" } private note ## The MC — Master of Ceremonies (AI announcer) Set emcee.voice + emcee.intro to script the show-opener, and guests[].emcee to script each guest's walk-on line — both spoken live by an AI announcer. The legacy guest_link is a receive-only audience door: an attendee can watch there and the host can privately invite them on stage, but it does not grant camera or microphone publishing by itself. For a scheduled guest, create a personal backstage URL with POST /api/v1/events/{slug}/guests/{guest_id}/access-link and send the returned backstage_url. Leave a script blank and it is auto-written at showtime. ## Reminders reminders toggles the automatic emails off the start time: h24 (24h before), h1 (1h before), live (going live), changed (details changed). Confirmation on sign-up is always on. ## Example curl -s -X POST https://api.flarea.ai/api/v1/events \ -H "Authorization: Bearer $FLAREA_API_KEY" \ -H "Content-Type: application/json" \ -d '{"title":"AI Risks Worth Talking About","brief":"Fireside with Marge Simpson, CISO at ACME, about AI risk. Cover: model risk nobody budgets for, war stories, audience hot seat.","starts_at":"2026-07-06T18:00:00Z"}' ## For agents Recommended setup in Claude Code / any CLI agent — add to your project notes: "Flarea manages my live events. API key in $FLAREA_API_KEY. Read https://api.flarea.ai/api/v1 for the API, then create/list/update events and pull attendee lists as asked." ## Sales intelligence — call analysis + coaching (Clincher) Capture, analyze, and search sales calls — and close ON the call. Auth: Authorization: Bearer cl_sk_… (mint at POST /api/v1/keys, revoke with DELETE /api/v1/keys?id=…). Credits: every workspace gets $10/mo of engine credits free; each call is priced in cents (analyze 4¢ · extract 2¢ · ask/followup/trackers 1¢ · rollup 4¢ · close FREE). 402 = CREDITS_EXHAUSTED. POST /api/v1/analyze transcript → summary, sentiment, talk-ratio, objections, MEDDIC/BANT scoring, risks, next steps, coaching (one call; auto-indexes for /ask) POST /api/v1/ask { question } → cited answer across the workspace's whole call history + knowledge base (RAG, honest "I don't know") POST /api/v1/followup { summary | transcript, contact_name?, tone? } → { subject, body } follow-up email POST /api/v1/close { title, line_items:[{name,qty,unit_price}], customer? } → deal + ready-to-sign proposal link (close ON the call) POST /api/v1/extract transcript → CRM fields (contact, company, deal value, products, next steps, stage) POST /api/v1/draft transcript → follow-up email or proposal outline (freeform cousin of /followup) POST /api/v1/trackers { trackers:[...] } → semantic mentions across all calls POST /api/v1/rollup { transcripts:[...] } → multi-call deal momentum + risk + next step POST /api/v1/calls/bot { meeting_url } → send a notetaker bot into the call GET /api/v1/usage this month's metered usage + remaining allowance (free) GET /api/v1/openapi.json machine-readable OpenAPI 3.1 for all of the above POST /api/mcp MCP server — the same primitives exposed as agent tools Quickstart (full sales-call intelligence in 20 lines): github.com/asafktz/webinar-show/blob/main/docs/API_QUICKSTART.md Reference app: github.com/asafktz/webinar-show/tree/main/examples/clincher-demo · LLM digest: /llms.txt