YouTube transcripts in more videos
GET /v1/raw/youtube/transcript/{video_id} and GET /v1/raw/youtube/channel-transcripts/{handle} now return a transcript when YouTube refuses the machine-translated track but serves the video’s own speech recognition. These requests previously answered 503 service_unavailable. The response reports that track as language: en-orig (or the requested code with -orig) and source: auto_generated.A language offered only as a machine translation that YouTube does not serve now returns 404 not_found instead of a retryable 503. The language parameter accepts caption language codes, such as en or pt-BR, or auto; wildcards and all return 400 validation_error.Missing Instagram usernames return 404 with posts too
GET /v1/raw/instagram/profile/{username} now returns 404 not_found for a username that does not exist when posts are requested (include_posts=true, the default), as it already did for profile-only lookups. These requests previously answered 503 service_unavailable. The response is not charged.Clearer errors for private profiles and temporary source refusals
Raw Instagram requests for a private account now return403 forbidden with error.details.upstream_error: profile_private and a message saying the account is private. Private TikTok accounts still return 403 forbidden and now carry the same profile_private detail. Previously the response was 403 scraping_failed, and MCP clients described it as a credential permission problem.Stop automatic retries when you see profile_private; the request fails the same way until the account is public. The response is not charged.When a source platform temporarily refuses a live request, raw endpoints now return a retryable 503 service_unavailable with error.details.upstream_error: upstream_unavailable, instead of 403 scraping_failed. For API-key and OAuth requests these responses include Retry-After; honor it and retry with bounded backoff.Raw Instagram and YouTube errors for x402 and MPP requests no longer include Retry-After or error.details.retry_after_seconds. Payment settles before the request runs, so a retry is a new payment; decide whether to retry yourself.Age-restricted YouTube videos still return 403 scraping_failed, now with a fixed message saying the video is age-restricted and error.details.retryable: false.Clearer errors for missing live targets
Two live endpoints no longer answer503 service_unavailable for a target that does not exist, so retry logic stops repeating a request that cannot succeed:GET /v1/raw/instagram/profile/{username}returns404 not_foundwhen the username does not exist.GET /v1/raw/youtube/transcript/{video_id}returns403 scraping_failedwitherror.details.upstream_error: content_unavailablewhen YouTube reports the video as unavailable. This does not confirm that the video was deleted.
More dependable campaign match results
POST /v1/creators/match results are more complete and consistent:- Creators that could not be analyzed are now marked with
match.analysis_unavailable: trueinstead of appearing as an ordinaryneutraldecision with a score of 0. They are still not charged API credits; retry them for a full analysis. Fewer creators end up in this state. match.decisionnow always follows the published score bands (goodat 0.70 and above,neutralfrom 0.50 to 0.69,avoidbelow 0.50), so decision and score never disagree for analyzed creators.- Match reasons can include
source_post_url, a link to the post identified bysource_post_id. - Every analyzed creator has at least one reason, and more
post_evidencereasons include a verbatimevidence_quote.
Faster creator matching and retryable timeouts
POST /v1/creators/match now scores large requests much faster, so batches up to the 100-creator limit complete reliably. Requests billed to API credits are no longer charged for creators that could not be analysed.Creator search, match and lookalike requests that cannot finish in time now return 503 service_unavailable with a Retry-After header instead of 500, so the SDKs and your retry logic can treat them as temporary. Search returns a ranked result in more of these cases instead of an error. See Error Handling.One MCP URL with sign-in
Connect Claude, Claude Code, Cursor, VS Code and other MCP clients tohttps://mcp.influship.com, then sign in when your client asks. During sign-in you choose whether research is billed to a pay-as-you-go API account or to your Influship app plan, so you no longer need to copy an API key into a config file. Headless setups can still send X-API-Key to the same URL, and lookalike discovery is available there alongside the creator workspace and live platform research. Existing /mcp configurations keep working. See the MCP Server guide.Runnable creator discovery examples
The cookbook now includes a complete Next.js creator search app, campaign comparison, and CSV export. The shortlist, lookalike and existing-list scoring recipes include complete bounded CLI workflows, expected output, costs and recovery guidance.The creator research app adds evidence review and saved API shortlists with no dependency installation. Download its v1.1.0 source archive.Fork the MIT example code from Influship examples or download the reviewed v1.0.0 source archives. Each recipe links its source, the TypeScript SDK guide, relevant API contracts and related workflows.Durable media URLs from GET /v1/posts
GET /v1/posts now returns durable hosted image URLs for media.url, media.thumbnail_url, and carousel_items[].thumbnail_url whenever a hosted copy is available. Previously these were platform links that expire within days. When no hosted copy exists, you still get the platform URL, which expires, so download it promptly. The response shape is unchanged.Live and raw endpoints under /v1/raw/ are unchanged: they return fresh platform URLs on each request.Live-data errors and ingest quota updates
Live and raw Instagram, YouTube, and TikTok endpoints now return503 service_unavailable in the standard error envelope whenever live data is temporarily unavailable. Previously some of these cases surfaced as a generic 500 internal_error or a non-standard body. Retry 503 responses with bounded backoff, honoring Retry-After when present; you are not charged for them. See Error Handling.POST /v1/creators/ingest no longer consumes your daily ingest quota when the handle does not exist (404) or cannot be validated right now (503). A slot is only used once the request is accepted. See Ingest Creators On Demand.Breaking raw Instagram profile contract update
GET /v1/raw/instagram/profile/{username} no longer returns highlight_reel_count, is_business, or is_professional. Remove reads of these properties from your integration.media_count is required but nullable. Handle null as an unavailable lifetime total; 0 means an observed empty profile. Do not substitute the number of returned posts for the lifetime total. Cached Instagram profiles and TikTok responses are unchanged.This changes the existing /v1 response contract. Update response validators and null handling before consuming the new response. See SDK guidance for client examples.YouTube discovery, pagination, and video details
YouTube search now supports upload-date, popularity, content-type, duration, country, and language filters. Responses include an opaquenext_cursor; pass it back as cursor to fetch another page. Search costs 0.5 credits per fetched page, independent of the number of results returned.Two new endpoints expand topic discovery and validation:GET /v1/raw/youtube/typeaheadreturns localized query suggestionsGET /v1/raw/youtube/video/{video_id}returns fresh views, likes, comments, duration, tags, and exact publication data when available
Automatic recovery for raw Instagram requests
Raw Instagram profile, post, and transcript requests now recover automatically from temporary source-platform interruptions. Confirmed missing posts, missing profiles, and private profiles continue to return their documented errors.Batch post and transcript requests now stop when your request expires or disconnects. Batch results remain in the same order as the requested shortcodes.Raw Instagram post & transcript reliability
Fixed errors affecting Instagram post lookups and video transcripts after a source-platform change. Individual post lookup, batch post lookup, and transcript endpoints now return consistent response shapes across the REST API, MCP tools, and no-code actors.Temporary source-platform failures now return retryable429/503 responses with Retry-After instead of a generic 500. Use the same backoff as other live-data requests — see Error Handling.Affected endpoints:GET /v1/raw/instagram/post/{shortcode}POST /v1/raw/instagram/postsGET /v1/raw/instagram/transcript/{shortcode}POST /v1/raw/instagram/transcripts
Raw Instagram post lookup and transcripts
You can now look up an individual Instagram post by shortcode and fetch video transcripts directly from the raw API. These endpoints return fresh raw post data, including fields that are not guaranteed on cached post-list responses, such as coauthors, paid partnership flags, tagged users, product mentions, display resources, video versions, music attribution, and location data.Available endpoints:GET /v1/raw/instagram/post/{shortcode}for one postPOST /v1/raw/instagram/postsfor up to 20 posts in one requestGET /v1/raw/instagram/transcript/{shortcode}for one video post transcriptPOST /v1/raw/instagram/transcriptsfor up to 10 video post transcripts in one request
post; call the post lookup endpoint separately when you need post metadata alongside a cached transcript.We’ve also added two Apify actors for no-code and workflow use cases:- Instagram Post Lookup
- Instagram Post Transcripts
MCP Server
Influship launched an MCP server with eight typed tools. For the current connection URLs, account setup, and tool catalog, see the MCP Server guide.Pay per request with x402 (no API key needed)
Anonymous AI agents can now call supported paid Influship endpoints via the x402 payment protocol. The agent makes a request without anX-API-Key, gets a 402 Payment Required response with a USDC-on-Base price, signs a payment, and retries without signup or billing setup.Pay with MPP (Stripe cards or USDC on Tempo)
The Machine Payments Protocol — Stripe and Tempo’s open standard for HTTP-native machine-to-machine payments — coexists with x402 on the same endpoints. Pay per request via Stripe SPT (Shared Payment Tokens, USD via card) or USDC on Tempo. Pick whichever rail your client supports; the API accepts both.Weighted lookalike seeds
Lookalike search now accepts up to 10 seeds with individual weights (0-1). Higher weights make the API lean more heavily on that seed’s characteristics when ranking results.This is particularly useful when you have campaign performance data — weight your best performers higher and the API finds more creators who skew toward that profile.Campaign match endpoint
POST /v1/creators/match scores creators against a campaign brief. Each result includes a decision (good, neutral, or avoid), a numeric score, and human-readable reasoning.The intent.query field (500 chars) describes the campaign. The optional intent.context field (2,000 chars) adds background — target demographics, content format preferences, brand guidelines.See Match Reasons for how to interpret the output.Credit-based rate limiting with trust tiers
Rate limits are now enforced using credit budgets instead of simple request counts. Heavier endpoints consume more budget than lighter ones. Trust tiers increase your limits based on lifetime spend — and never downgrade.See Rate Limits & Tiers for the full tier table.TypeScript SDK
The official TypeScript SDK is available on npm. It provides typed methods, request/response models, and error classes likeRateLimitError and APIError.