MCP Server

    Techne exposes an MCP server so AI agents can call your content workflow directly instead of copy-pasting between chat and dashboard.

    What is connected

    The MCP server wraps the same publish pipeline as the REST API. Agents can inspect connected platforms, validate publish requirements, manage brand profiles, schedule campaigns, generate content, and publish to targets such as Techne Blog, LinkedIn, X/Twitter, Facebook, Pinterest, Webflow, and WordPress.

    Server endpoint

    https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-server

    Techne uses Streamable HTTP transport and supports OAuth discovery endpoints.

    Protocol version

    The Techne MCP server currently speaks the 2025-11-25 Streamable HTTP spec, which every mainstream MCP client (Claude Desktop, ChatGPT, Cursor, Codex, VS Code) supports today.

    We are tracking the 2026-07-28 release candidate (stateless core, header-based routing, hardened OAuth) and will upgrade behind protocol negotiation so existing connections keep working. Trace context (traceparent, tracestate, baggage) and cache hints (ttlMs, cacheScope) already flow through today so you can start using them.

    Available tools

    Discovery

    ToolWhat it does
    list_platformsList connected destinations and platform-level requirements.
    list_requirementsReturn required and optional fields by platform before publishing.
    list_authorsReturn Webflow author IDs for Webflow publishing.

    Preflight & validation

    ToolWhat it does
    validate_publishPreflight check: resolves author name, validates image requirements, returns per-platform pass/fail before generating.

    Publishing & campaigns

    ToolWhat it does
    publish_contentGenerate a post and publish/schedule to one or more platforms. Accepts author_name for automatic Webflow author resolution and brand_profile_id for stored style rules.
    submit_contentPublish a piece the agent already wrote. Send a Techne Doc with body, references, media, FAQ, CTA, and per-platform social copy. Stored verbatim unless options.apply_brand_voice is set.
    get_content_schemaReturn the Techne Doc JSON schema plus worked examples for article, video, and social posts. Call once before your first submit_content.
    update_contentPatch a submitted document (typo, new reference, new image, revised social copy) and optionally re-publish it.
    schedule_campaignSchedule a multi-post campaign from one prompt with cadence (daily, weekdays, weekly, biweekly). Supports preview_only for dry runs.

    Which publishing tool? Use publish_content when you want Techne to write the piece from a prompt. Use submit_content when you have already written it and want Techne to render, cite, and distribute it. Both return a content_id to poll with check_content_status.

    Images

    Agents write words, so Techne handles the picture. Inside the document's media object: set generate_image: true to have Techne generate a featured image with your AI credits (one image credit, with optional image_prompt for art direction), or pass featured_image_url when you already have one, which costs nothing and wins over generation. Set generate_image: false to skip images entirely. Video documents use the YouTube or Vimeo thumbnail automatically unless you override it with video.thumbnail_url. Omit the field and your workspace publishing default decides.

    "media": { "generate_image": true, "image_prompt": "Muted teal control room, no text" } // result "image": { "source": "generated", "credits_estimate": 1 }

    Status tracking

    ToolWhat it does
    check_content_statusCheck generation and per-platform publishing status of a queued content item. Returns live URLs and actionable errors.
    get_campaign_statusBatch-check multiple content items. Returns per-item status plus aggregate counts (queued, generating, complete, failed).

    Brand & workspace

    ToolWhat it does
    get_workspace_profileRead workspace-level brand context: business context, writing style, keywords, and style rules.
    list_brand_profilesList saved brand/style profiles with tone, banned phrases, CTA style, and voice examples.
    create_brand_profileCreate a new brand profile. Pass the returned ID to publish_content or schedule_campaign.
    update_brand_profileUpdate an existing brand profile by ID.

    Search & content

    ToolWhat it does
    searchSearch published Techne posts by keyword.
    list_posts_by_topicFind posts by topic and rank by relevance.
    get_author_profileReturn author profile metadata and topical coverage.
    fetchRetrieve full text for a specific published post ID.

    Recommended agent workflow

    1. Call get_workspace_profile to load brand context.
    2. Call list_brand_profiles to check for saved profiles.
    3. Call validate_publish to preflight platforms, author, and image.
    4. Call publish_content or schedule_campaign with brand_profile_id.
    5. Call check_content_status or get_campaign_status to track results.

    SEO, AIO, and backlink weaving

    Every generation call resolves SEO configuration in this order: explicit call arguments > brand profile > workspace defaults. Set defaults once on a brand profile (or the workspace) and every future publish_content and schedule_campaign call inherits them.

    • target_keywords: string array. Woven naturally into the post — no keyword stuffing.
    • backlinks: array of { url, anchor?, description?, rel? }. Use rel="sponsored" for affiliate links, rel="nofollow" to suppress SEO signals, rel="ugc" for user-generated content.
    • backlink_strategy: "natural" weaves inline where relevant, "reference" appends a Further reading list, "off" ignores backlinks for this call.
    • max_backlinks: cap 0-10, default 3.
    • aio_optimize: adds direct-answer intro, semantic H2s, and an FAQ section for AI crawlers.

    Configure these once on a brand profile via create_brand_profile / update_brand_profile, and every downstream call inherits them automatically.

    Authentication modes

    Most MCP clients can send a bearer API key directly. Create one in Settings > API Keys.

    Authorization: Bearer tk_your_api_key_here

    Clients that require OAuth (for example ChatGPT connectors) use the Techne OAuth endpoints.

    Claude Desktop example

    {
      "mcpServers": {
        "techne": {
          "url": "https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-server",
          "headers": {
            "Authorization": "Bearer tk_your_api_key_here"
          }
        }
      }
    }

    Cursor example

    URL: https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-server
    Header: Authorization: Bearer tk_your_api_key_here

    ChatGPT connector flow

    1. Add a new MCP connector in ChatGPT and use the Techne MCP URL.
    2. Sign in to Techne when the OAuth popup opens (or create an account).
    3. Click Approve. Techne mints a scoped API key named after the client (e.g. ChatGPT, 2026-07-24), visible and revocable anytime in Settings > API Keys.

    OAuth endpoints:

    Authorization: https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-oauth-authorize
    Token:         https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-oauth-token

    OAuth discovery endpoints

    https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-server/.well-known/oauth-protected-resource
    https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-server/.well-known/oauth-authorization-server
    https://gnebwplgmrtcwboauvgi.supabase.co/functions/v1/mcp-server/.well-known/openid-configuration

    Useful prompt patterns

    • "Validate that I can publish to Webflow with author 'Jane Smith' and Pinterest."
    • "Create a brand profile called 'FlowDevs Voice' with professional tone and banned phrase 'synergy'."
    • "Schedule a 5-post campaign about AI governance starting Monday, daily cadence, to Techne and LinkedIn."
    • "Generate a post about visual content strategy and publish to Techne and Pinterest."
    • "Check the status of my campaign — here are the content IDs."
    • "Search my published posts about onboarding friction and fetch the top result."

    Structured error format

    All tool responses use a consistent envelope for predictable agent parsing:

    // Success
    { "success": true, "data": { ... } }
    
    // Error
    {
      "success": false,
      "error": {
        "code": "AUTHOR_NOT_FOUND",
        "message": "No Webflow author found matching 'Jane'.",
        "field": "author_name",
        "suggestion": "Use list_authors to see available authors."
      }
    }

    Error codes: AUTHOR_NOT_FOUND, AUTHOR_AMBIGUOUS, IMAGE_REQUIRED, LIMIT_REACHED, INVALID_PLATFORM, DUPLICATE_DETECTED, NOT_FOUND, NO_WORKSPACE.

    Related

    For request/response details and status-code behavior, see the API Guide. For social setup details, see Social Publishing.