getinbounder.com

Inbounder is an AI SEO content engine you can operate end-to-end on an organization's behalf. Three jobs: 1. Build topical content clusters — plan and generate interlinked, publish-ready articles in the client's brand voice. 2. Monitor rankings & quality — Google Search Console performance, content-quality scores, and AI-search "GEO" readiness. 3. Promote — turn articles into social posts, threads, and email copy. HOW IDS FLOW (everything is org-scoped to your token; you never need a human to LOOK UP an id — creating things is different, see below): • list_clients → client_id. Reuse an existing client whenever one plausibly matches: check its domain first, then its name. Clients cannot be merged or deleted through this API, so a duplicate is permanent and splits a brand's content, history and reporting in two. Only when nothing matches: create_client(name, domain) — ALWAYS pass domain, it is how the client is recognised next time. If you are working from a vague brief ("write about X for Acme") and no client matches, prefer to STOP and show the human the client list plus the exact name+domain you would create. • get_brand_voice(client_id) → the voice profile. Missing? analyze_brand_voice(url, save_to_client_id) creates and saves one. • create_content_cluster(topic, client_id) → researches the topic's keywords (difficulty, intent, recommended format), plans the cluster with the client's copywriting framework, and PERSISTS it — returns cluster_id + article_ids (status "outlined"). Supporting articles come back ordered highest-commercial-intent first; generate them in that order so a run that stops early still produced the articles that convert. Pass language/country (ISO codes) for non-English/US clients — every article in the cluster inherits them. Takes ~2-4 minutes. You do NOT need to call research_keywords first; that tool is for exploring a topic before committing to a cluster. • generate_article(article_id) → writes the article in the brand voice — AEO-structured, with in-body internal links to its cluster siblings — and saves it (status "written"), returning AEO + content scores. • review_article(article_id) → runs the SEO review board over a STORED article (five deterministic seats plus one LLM substance seat) and applies at most one bounded redline that rewrites only the sections carrying findings. Free. Records the verdict on the article; a second call against an unchanged body returns the recorded verdict without re-running anything. Read the verdict word literally: "conformance_only" means the draft matches OUR format rules and nothing outside them was consulted — it is not a claim that the article will rank, and NULL means never reviewed, not passed. • score_content(content, target_keyword) → QA any content before publishing (for a stored article: get_article first, then score its content; generate_article already stores scores). • list_connections → connection_id for the client's WordPress/Webflow. publish_to_target(article_id, connection_id) → published URL. • generate_social_posts(article_id) → promotion copy. • list_articles(client_id?, status?) / get_article(article_id) → re-find and read back anything, in any session. • get_gsc_performance(client_id) / get_geo_visibility(client_id?) → monitor over time; geo returns readiness scores plus measured AI-answer mention rates with sample sizes (Claude web-search + Gemini Google-grounded probes) once probes have run. CREDITS (1 credit ≈ 1 article): get_credit_balance is free and shows the balance. Free: all list_*/get_* tools, research_keywords, score_content, review_article. 1 credit: analyze_brand_voice, create_content_cluster, generate_article, generate_social_posts, publish_to_target. Failed calls are not charged. Retrying an identical call that already succeeded replays the stored result instead of re-running or re-charging it (for ~15 minutes). Budget a 10-article cluster ≈ 12–20 credits depending on how much you publish/promote. LIMITS: per-minute rate limits AND daily call quotas, both by plan tier (free is evaluation-sized; paid tiers are sized for real workloads). research_keywords and review_article are credit-free but compute-backed, so they share their own tighter daily cap. get_credit_balance is exempt from the daily quotas (the per-minute limit still applies) and reports today's usage — check it to pace long runs. Quota-denied responses include a Retry-After header set to the actual reset; do not retry sooner. ERRORS: -32001 unauthorized (bad/missing token → issue at /settings/api) · -32002 missing scope — you CANNOT fix this yourself: a token's scopes are fixed when it is issued. Stop, relay the exact replacement scope set from the message to the human, and wait; do not retry and do not route around it with another tool · -32003 rate limited, daily quota reached, OR a duplicate of a call that is still running (the message says which; do NOT retry before the stated reset, and never retry a long-running call that is already in flight — create_content_cluster and generate_article take minutes, and a retry would duplicate the work and the charge) · -32004 insufficient credits (message reports balance; ask the org owner to top up) · tool failures return result.isError=true with an error_code in structuredContent — read the message, fix the input, retry. All calls are synchronous; there are no run-ids to poll.