Socialneuron

Delete an owned carousel content-history record from one project. Stored media is retained until the normal retention cleanup; this does not delete already-published platform posts.

Delete a YouTube comment (YouTube only today). Only works for comments owned by the authenticated channel.

Permanently delete an owned content plan in one project. This does not cancel posts that were already scheduled from the plan.

Detect significant performance changes: spikes, drops, viral content, trend shifts. Compares current period against previous equal-length period. No AI call, no credit cost.

Short description — what the tool does AND when the model should call it. Include concrete triggers ("use when the user asks about X").

Execute a recipe template with the provided inputs. This creates a recipe run that processes each step sequentially. Long-running recipes will return a run_id you can check with get_recipe_run_status. Recipes containing a distribution step (publishing or webhook egress) additionally require the mcp:distribute scope.

Explains what brand data is available vs missing for a project. Returns a human-readable summary of completeness, confidence levels, and recommendations for improving the brand profile.

Export brand palette and typography as design tokens. Supports CSS custom properties, Tailwind config, and Figma Tokens JSON formats.

Analyze a website URL and extract brand identity data including brand name, colors, voice/tone, target audience, and logo. Uses AI-powered analysis of the page HTML. Useful for understanding a brand before generating content for it.

Extract text content from any URL — YouTube video transcript + metadata (title/views/channel), or article text, or product page features/benefits/USP. YouTube URLs auto-route to transcript+metadata extraction; channel URLs return a recent-video list. Use before generate_content to repurpose existing content, or before plan_content_week to base a content plan on a source URL.

Fetch a public Social Neuron knowledge document by ID. Use IDs returned by the search tool.

Get project-scoped post performance metrics — views, likes, comments, shares, and engagement rate. Filter by platform, time range (default 30 days), or specific content_id. Call refresh_platform_analytics first if data seems stale. Results sorted by most recent capture.

Get current trending topics for content inspiration. Source "youtube" returns trending videos with view counts, "google_trends" returns rising search terms, "rss"/"url" extracts topics from any feed or page. Results cached 1 hour — set force_refresh=true for real-time. Feed results into generate_content or plan_content_week.

Fetch YouTube channel analytics. Supports channel overview, daily breakdown, video-specific metrics, and top-performing videos. Requires a connected YouTube account.

Find optimal posting time slots for one brand/project based on preferred posting times and that project's existing schedule. Returns non-conflicting slots sorted by engagement score.

Find QA-gated high-performing short-form videos in the project's niche. Returns extracted hook patterns, content structures, and pre-compiled replication prompts you can use to generate new content on a different topic. Only QA-gated winners with a populated replication prompt are returned.

Generate carousel slide content (headlines, body text, emphasis words per slide). Supports Hormozi-style authority format and educational templates. Returns structured slide data — render visually then publish via schedule_post with media_type=CAROUSEL_ALBUM and 2-10 media_urls on Instagram.

Create a script, caption, hook, or blog post tailored to a specific platform. Pass project_id to auto-load brand profile and performance context, or call get_ideation_context first for full context. project_id may be omitted only when exactly one project is accessible; multi-project and zero-project accounts fail closed before generation. Output is draft text ready for quality_check then schedule_post.

Description is built at runtime.

Generate a performance summary for a time period. Includes metrics, trends vs previous period, top/bottom performers, platform breakdown, and actionable recommendations. No AI call, no credit cost.

Start an async AI video generation job — returns a job_id immediately. Poll with check_status every 10-30s until complete. Available models and accepted options are defined by the live input schema. Credit usage varies by model, duration, audio, and resolution; use get_credit_balance to confirm sufficient balance before starting. The accepted job response reports the credits charged.

Generate a voiceover audio file for video narration. Returns an R2-hosted audio URL. Use after create_storyboard to add narration to each scene, or standalone for podcast intros and ad reads. Pass project_id to keep the asset with the correct brand/project. Costs 15 credits per generation. If a prior call returned "still generating", pass its task_id back as resume_task_id to check again WITHOUT a second charge — never call this fresh to "retry" a still-generating job.

Get autopilot system overview: active config count, recent execution results, credits consumed, and next scheduled run time. Use as a dashboard check before modifying autopilot settings.

Analyze project-scoped post analytics data to find the best times to post for maximum engagement. Returns the top 5 time slots (day of week + hour) ranked by average engagement.

Load the active persisted brand profile for a project from brand_profiles.

Fetches a project's 4-layer brand runtime: messaging (value props, pillars, proof points), voice (tone, vocabulary, blocked terms), visual identity (palette, typography, composition), and audience details (archetype, target). Includes extraction confidence scores.

Check how much of the per-session budget has been consumed. Tracks credits spent and assets created in this MCP session against configured limits. Use to avoid hitting budget caps mid-workflow.

Load a persisted content plan by its UUID — returns the full plan including all posts, scheduling status, and approval state. Use to inspect a plan before update_content_plan or schedule_content_plan. plan_id comes from save_content_plan, plan_content_week (when persisted), or list_plan_approvals. For just the approval state, list_plan_approvals is cheaper.

Description is built at runtime.

Get synthesized ideation context from performance insights. Returns the same prompt-injection context used by ideation generation.

Get a one-call dashboard summary of the feedback loop state (brand profile, recent content, and current insights).

Get your MCP API usage breakdown for the current billing month. Shows per-tool call counts and credit usage. Useful for monitoring API consumption and staying within tier limits.

Get a fresh download URL for stored media when a previous URL has expired. Pass the r2_key from upload_media or check_status, or the job_id from check_status.

Query project-scoped performance insights derived from post analytics. Returns metrics like engagement rate, view velocity, and click rate aggregated over time. Use this to understand what content is performing well.

Check status of a pipeline run, including stages completed, pending approvals, and scheduled posts.

Get full details of a recipe template including all steps, input schema, and estimated costs. Use this before execute_recipe to understand what inputs are required.

Check the status of a running recipe execution. Shows progress, current step, credits used, and outputs when complete.

Fetch the full body of a single Social Neuron skill by slug — the hand-maintained strategy/specs plus the machine-maintained "what's working now" compiled section. Use this after list_skills when the user wants the actual playbook for a platform (e.g. "show me the TikTok skill"). Returns the skill body, compiled section, tier, and linked recipe slug (if any).

List autopilot configurations showing schedules, credit budgets, last run times, and active/inactive status. Use to check what is automated before creating new configs, or to find config_id for update_autopilot_config.

List YouTube comments (YouTube only today) — pass video_id (11-char string, e.g. "dQw4w9WgXcQ") for a specific video, or omit for recent comments across all channel videos. Returns comment text, author, like count, and reply count. Use page_token from previous response for pagination.

List all available Remotion video compositions defined in Social Neuron. Returns composition IDs, dimensions, duration, and descriptions. Use this to discover what videos can be rendered with render_demo_video.

Check which social platforms have active OAuth connections for posting. Call this before schedule_post to verify credentials. Pass project_id to list the accounts for a specific brand/project, then pass the returned account id as account_id/account_ids when posting. If a platform is missing or expired, the user needs to reconnect at socialneuron.com/settings/connections.

List the curated subset of pre-built Hyperframes blocks (transitions, social overlays, data-viz, branding, decorative) the agent can compose into an HTML video composition. Returns block IDs with categories + 1-line descriptions. No network call — returns a static catalog.

List approval items for a content plan, optionally filtered by status (pending / approved / rejected / edited). Use to check what needs review before scheduling, or to audit decisions after the fact. plan_id comes from get_content_plan or save_content_plan. For a single item's full state, get the plan via get_content_plan instead — that includes per-post approval data inline.

List recent published and scheduled posts with status, platform, title, and timestamps. Use to check what has been posted before planning new content, or to find post IDs for fetch_analytics. Filter by platform or status to narrow results.

List available recipe templates. Recipes are pre-built multi-step workflows like "Weekly Instagram Calendar" or "Product Launch Sequence" that automate common content operations. Use this to discover what recipes are available before running one.

List Social Neuron content workflow skills available to the authenticated user. A skill is a brand-locked multi-step pipeline (research → hook → script → visuals → voice → captions → assembly → quality gate → schedule) inspired by documented viral patterns (MrBeast 3-second hook, Hormozi pattern interrupt, etc.). Use this tool when the user asks "what can SN do", "what skills are available", "show me viral templates", or before calling run_skill so you can pick the right one.

Moderate a YouTube comment on your channel (YouTube only today) — set status to "published" (approve) or "rejected" (hide from public view but kept in moderation queue). Use after list_comments surfaces a comment that needs action. For permanent removal use delete_comment instead. comment_id comes from list_comments results.

Open a project-scoped performance dashboard with views, engagement, platform mix, and top-post metrics. Use it to review results visually before planning the next content cycle.

Open a project-scoped interactive calendar for the current week. Users can filter, inspect, quick-create, suggest a slot, and reschedule pending posts with optimistic conflict protection.

Generate a full content plan with platform-specific drafts, hooks, angles, and optimal schedule times. Pass a topic or source_url — brand context and performance insights auto-load via project_id. Output feeds directly into quality_check_plan then schedule_content_plan. Costs ~5-15 credits depending on post count.

Post a new top-level comment on a YouTube video (YouTube only today), authored as the connected channel. Use for proactive engagement on your own videos. For replies to existing comments use reply_to_comment instead — this tool only creates top-level comments. video_id comes from list_recent_posts (platform_post_id field) or any YouTube URL (the v= parameter, 11 chars). Subject to YouTube anti-spam rate limits; calls return rate_limited if exceeded.

Score post quality across 7 categories: Hook Strength, Message Clarity, Platform Fit, Brand Alignment, Novelty, CTA Strength, and Safety/Claims. Each scored 0-5, total 35. Default pass threshold is 26 (~75%). Run after generate_content and before schedule_post. Include hashtags in caption if they will be published — they affect Platform Fit and Safety scores.

Batch quality check all posts in a content plan. Returns per-post scores and aggregate pass/fail summary. Use after plan_content_week and before schedule_content_plan to catch low-quality posts before publishing.

Queue analytics refresh jobs for posts from the last 7 days in one project across its connected platforms. Call this before fetch_analytics if you need fresh data. Returns immediately — data updates asynchronously over the next 1-5 minutes.

Render a Remotion composition to an MP4 or GIF file locally. Uses the Remotion bundler and renderer from the root project. This can take 30-120 seconds depending on composition length. Output is saved to public/videos/.

Render an HTML video composition (Hyperframes) to MP4 — frame-accurate, no React build step. The page MUST expose window.__hf = { duration: <seconds>, seek: (t) => void }; the renderer calls seek(t) per frame (GSAP timelines work when driven from seek). Missing window.__hf fails after the render timeout and is auto-refunded. Use list_hyperframes_blocks for the pre-built block catalog. Pass project_id to keep the render with the correct brand/project. Returns a job ID — poll with check_status.

Render a Remotion template video in the cloud. Creates an async render job that is processed by the production worker, uploaded to R2, and tracked via async_jobs. Returns a job ID that can be polled with check_status. Costs credits based on video duration (3 base + 0.1/sec). Use list_compositions to see available template IDs. Pass project_id to keep the render (and its R2 storage path / brand context) scoped to the correct project. Omission auto-resolves only when exactly one project is accessible; zero-project and multi-project accounts fail closed before spend.

Reply to a YouTube comment (YouTube only today). Get the parent_id from list_comments results. Reply appears as the authenticated channel. Use for community engagement after checking list_comments for questions or feedback.

Move an existing pending or scheduled post to a new future time without creating a duplicate. Pass project_id for the post's brand. expected_scheduled_at is recommended: it prevents overwriting a change made in another client after the calendar was loaded.

Approve, reject, or edit a single pending plan approval item. Use to act on items surfaced by list_plan_approvals. decision="edited" REQUIRES edited_post containing the modified post fields — passing "edited" without edited_post returns an error. Once decided, an item cannot be re-decided (immutable transition). reason is optional but recommended for "rejected" or "edited" to leave a paper trail. After all items are decided, schedule_content_plan publishes only the approved (and edited) ones.

Run the full content pipeline: research trends → generate plan → quality check → auto-approve → schedule posts. Chains all stages in one call for maximum efficiency. Set dry_run=true to preview the plan without publishing. To schedule posts, set schedule_confirmed=true after the user explicitly approves publishing. Check check_pipeline_readiness first to verify credits, OAuth, and brand profile are ready.

Preview only: returns a structured run preview, estimated credits, and a dashboard deep link — it does not execute the skill. Run a Social Neuron workflow skill end-to-end (brand-locked content production). Returns a structured run preview with the exact step plan, credit cost, and a deep-link to launch the run in the SN dashboard. A future release executes in-process so you can stream step-by-step progress back to the user. Call list_skills first to discover available skill ids.

Save (or replace) the active brand profile for a project — voice, target audience, content pillars, claims, etc. Use after extract_brand has produced a draft AND the user has reviewed it, or when the user explicitly edits the profile. brand_context is the full profile payload from extract_brand or get_brand_profile. project_id defaults to the active project context. Overwrites the previous active profile (one per project) — pass the complete profile, no merge semantics. Use change_summary to leave an audit trail.

Save a content plan to the database for team review, approval workflows, and scheduled publishing. Creates a plan_id you can reference in get_content_plan, update_content_plan, and schedule_content_plan.

Schedule all posts in a content plan. Optionally auto-assigns time slots and runs quality checks before scheduling. Supports dry-run mode.

Publish or schedule a post to connected social platforms. ALWAYS call `list_connected_accounts` FIRST — if the target platform is not connected, call `start_platform_connection` to get a one-time browser deep link the user opens to complete the platform OAuth (this is a one-time setup on socialneuron.com, not another OAuth in Claude). After they approve, call `wait_for_connection` and only then call schedule_post. For Instagram carousels: use media_type=CAROUSEL_ALBUM with 2-10 media_urls. For YouTube: title is required. schedule_at uses ISO 8601 (e.g. "2026-03-20T14:00:00Z") — omit to post im…

Search public Social Neuron product, integration, developer, and MCP tool knowledge. Uses the standard ChatGPT MCP search schema and never returns private account content.

Search available tools by name, description, module, or scope. Use "name" detail (~50 tokens) for quick lookup, "summary" (~500 tokens) for descriptions, "full" for complete input schemas. Start here if unsure which tool to call — filter by module (e.g. "planning", "content", "analytics") to narrow results.

Begin connecting a social platform (Instagram, TikTok, YouTube, etc.). Returns a single-use deep link the user opens in a browser to complete the one-time OAuth handshake on socialneuron.com. This is NOT another OAuth in Claude — platform connections require a browser session because the social platforms (Meta, Google, TikTok) only accept callbacks on socialneuron.com. After the user clicks the link and approves on the platform, call `wait_for_connection` to detect completion. Link expires in 2 minutes; mint a new one if needed. Use `list_connected_accounts` first to check whether the platform…

Create pending approval items for each post in a plan and mark plan status as in_review.

Suggest next content topics based on performance insights, past content, and competitor patterns. No AI call, no credit cost — purely data-driven recommendations.

Update an existing autopilot configuration. Can enable/disable, change schedule, or modify credit budgets.

Edit individual posts in a persisted content plan — change caption, title, hashtags, hook, or angle. Use after get_content_plan when the user wants to revise drafts before scheduling. Each post_updates entry must include post_id from the loaded plan; only the fields you pass get updated, others stay as-is. Does NOT trigger publishing — call schedule_content_plan separately when ready.

Update platform-specific voice overrides (samples, tone/style, CTA/hashtag strategy).

Upload media to durable storage. Returns an r2_key that can be passed to schedule_post. Three input modes: (1) local file path (stdio mode only), (2) public URL fetched by the server, (3) inline base64 via file_data (remote agents, ≤10MB decoded). AGENT ROUTING GUIDE: If the media was produced by another tool here (generate_image, generate_video, create_carousel, etc.), use the returned job_id or r2_key directly with schedule_post — do NOT download and re-upload. For user-authored files larger than ~1MB, ask the user to upload them in the Social Neuron app (Assets) so bytes never flow through …

Read the per-layout field constraints used by the visual QA gate (font size, effective width, max lines per field). Useful when generating slide text — lets you produce content that fits the first time. Returns the same TEMPLATE_FIELD_CONSTRAINTS table used at publish-time.

Run a pre-render visual QA check on carousel slides before publishing. Predicts text overflow against the per-layout font-size and container constraints of the rendering templates. Does NOT call Gemini Vision OCR (that runs inside the worker on the rendered PNG). Use after generate_carousel and before schedule_post to catch clipped text and single-line overflows. Spellcheck is skipped at the MCP layer — the worker handles it with the brand vocab allowlist.

Poll until a platform connection becomes active. Use after `start_platform_connection` while the user completes the browser OAuth flow. Returns when the account row appears with status=active, or when the timeout elapses. Default timeout 120s, max 600s.