Every call, parameter, result, and side effect.

Read the authenticated creator's role, services, audiences, proof, CTA, language, and normalized reference text.

How it works

Reads the authenticated account's normalized Expertise profile. If the profile row does not exist yet, GuideBlink creates an empty default row before returning it.

Returns

profile (id, role, services, target_clients, audience_targets, proof_points, CTA, output language, writing samples, extracted reference text, timestamps) and edit_url.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only; no write approval.

Parameters

No parameters. Send an empty arguments: {} object.

{
  "jsonrpc": "2.0",
  "id": "call-get_expertise_profile",
  "method": "tools/call",
  "params": {
    "name": "get_expertise_profile",
    "arguments": {}
  }
}

Patch only the supplied expertise fields. This is a free manual save and does not use AI or credits.

How it works

Patches only the fields supplied in arguments and returns the complete saved profile. audience_targets is normalized to at most three unique readers and also updates target_clients; supplying target_clients instead rebuilds audience_targets. An empty supplied text clears that field, except an empty output_language becomes English.

Returns

profile, edit_url, updated_fields, and credits_charged: 0.

Credits

0 credits; this is a manual save and does not extract or generate text with AI.

Approval / side effect

Account write; Codex should show the proposed profile changes before saving.

• cta_url must be an allowed public HTTP or HTTPS URL.
• Unknown fields are rejected.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-update_expertise_profile",
  "method": "tools/call",
  "params": {
    "name": "update_expertise_profile",
    "arguments": {
      "role": "AI automation engineer",
      "services": "Production AI agents and workflow automation",
      "audience_targets": [
        "SaaS founders",
        "Operations leaders"
      ],
      "proof_points": "Describe only verified projects and outcomes.",
      "output_language": "English"
    }
  }
}

Read the creator's hook, post, and CTA examples plus writing rules and advanced instructions.

How it works

Reads the authenticated account's current hook examples, full Post examples, CTA examples, writing rules, and advanced instructions. It creates an empty default Voice row when needed.

Returns

voice (id, five editable Voice fields, timestamps) and edit_url.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only; no write approval.

Parameters

No parameters. Send an empty arguments: {} object.

{
  "jsonrpc": "2.0",
  "id": "call-get_voice_profile",
  "method": "tools/call",
  "params": {
    "name": "get_voice_profile",
    "arguments": {}
  }
}

Patch only the supplied Voice fields. This is a free manual save and does not analyze examples or spend credits.

How it works

Patches only supplied Voice fields, preserving every omitted field. Supplying an empty string clears that field. This call stores examples and rules exactly as manual Voice input; it does not run the separate AI extraction workflow.

Returns

voice, edit_url, updated_fields, and credits_charged: 0.

Credits

0 credits; no LLM call.

Approval / side effect

Account write; show the exact proposed text before saving.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-update_voice_profile",
  "method": "tools/call",
  "params": {
    "name": "update_voice_profile",
    "arguments": {
      "hook_examples": "Most agent evals start too late.",
      "cta_examples": "Comment EVALS and I will send the guide.",
      "writing_rules": "Use concrete examples. Do not invent metrics."
    }
  }
}

List owned editable content templates. Select one of these template IDs before starting generation.

How it works

Ensures the account's ten default templates exist, then lists owned editable templates. kind can filter the result; omit it to receive both Post and Guide templates.

Returns

templates with id, kind, name, description, structure, instructions, source, and updated_at; plus count and edit_url.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only; no write approval.

• Use a returned Guide template_id with create_guide. Post template_id is optional.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-list_content_templates",
  "method": "tools/call",
  "params": {
    "name": "list_content_templates",
    "arguments": {
      "kind": "guide"
    }
  }
}

Return the current model-dependent credit reservation and whether the account can afford it. No credits are spent.

How it works

Calculates the conservative credit reservation for the requested content kind using the models currently selected on the Voice page. It also reads the current credit balance and affordability without reserving anything.

Returns

kind, quoted_credits, available_credits, can_generate, pricing_url, selected models, and the reservation/refund note.

Credits

0 credits; quoting does not reserve or spend credits.

Approval / side effect

Read-only. Call this immediately before create_post or create_guide and show the quote to the user.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-get_generation_quote",
  "method": "tools/call",
  "params": {
    "name": "get_generation_quote",
    "arguments": {
      "kind": "post"
    }
  }
}

Reserve the quoted credits and queue three editable LinkedIn drafts. Call the quote tool and obtain user approval first.

How it works

Validates the brief, reserves the current quoted credits atomically, creates a Post project, and queues one background job that produces three drafts. It returns immediately; poll get_generation_job. Reusing the same request_id for a Post returns the original job and does not reserve credits twice.

Returns

project_id, job_id, kind, status, progress, quoted_credits, available_credits, request_id, duplicate_request, editor_url, pricing_url, and poll_with.

Credits

Reserves the quoted amount. After completion, unused reserved credits are returned based on measured model cost. A safe failure settles or refunds the reservation according to recorded usage.

Approval / side effect

Paid write; call get_generation_quote and obtain explicit user approval first.

• source_urls accepts at most 12 safe public URLs.
• If template_id is supplied, it must identify an owned Post template.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-create_post",
  "method": "tools/call",
  "params": {
    "name": "create_post",
    "arguments": {
      "request_id": "post-agent-evals-001",
      "topic": "Why final-answer grading is not enough for production AI agents",
      "title": "Agent evals beyond the final answer",
      "audience_targets": [
        "AI engineering leaders",
        "Product managers"
      ],
      "use_voice": true,
      "options": {
        "goal": "Teach one practical evaluation lesson",
        "cta": "Invite readers to request the guide"
      }
    }
  }
}

Reserve the quoted credits and queue a researched guide. Requires an owned guide template; companion posts are not generated automatically.

How it works

Validates the brief, reserves the quote, creates a Guide project, and queues the researched long-form generation job. A Guide template is mandatory. To expand a Post, supply both source_project_id and source_post_variant. Companion Posts are not generated by this call. Reusing the same Guide request_id returns the original job without another reservation.

Returns

project_id, job_id, kind, status, progress, quoted_credits, available_credits, request_id, duplicate_request, editor_url, pricing_url, and poll_with.

Credits

Reserves the quoted Guide amount; unused reserved credits are returned after measured usage. Failed work is settled or refunded from recorded usage.

Approval / side effect

Paid write; quote and explicit user approval are required first.

• Only one active Guide generation is allowed per user.
• source_post_variant without source_project_id is rejected.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-create_guide",
  "method": "tools/call",
  "params": {
    "name": "create_guide",
    "arguments": {
      "request_id": "guide-agent-evals-001",
      "topic": "A practical system for evaluating production AI agents",
      "title": "A Complete Guide to AI Agent Evals",
      "audience_targets": [
        "AI engineering leaders"
      ],
      "template_id": 12,
      "source_urls": [
        "https://platform.openai.com/docs/guides/evals"
      ],
      "options": {
        "must_include": "Outcome grading, trajectory grading, and repeated reliability"
      }
    }
  }
}

Poll one owned generation job. Completed Post jobs include all drafts; completed Guide jobs include Markdown and research sources.

How it works

Reads one owned job and its current state. While work is active, poll_after_seconds is 2. A completed Post includes all three drafts; a completed Guide includes Markdown and citation sources. A failed job includes its safe error and final charged amount.

Returns

Job identity, kind, status, progress, reservation/charge fields, provider/model, error, timestamps, current balance, editor/pricing URLs and poll_after_seconds; plus drafts for a completed Post or guide_markdown, sources, and public_url for a completed Guide.

Credits

0 credits for polling.

Approval / side effect

Read-only; no write approval.

• Typical states are queued, researching, drafting, validating, completed, and failed.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-get_generation_job",
  "method": "tools/call",
  "params": {
    "name": "get_generation_job",
    "arguments": {
      "job_id": 321
    }
  }
}

List owned content projects so a client can find the project ID and available Post variants before editing settings or creating media.

How it works

Lists owned Post and Guide projects, newest updated first. Use it to resolve a project_id and to discover which Post draft variants exist before a visual export.

Returns

projects with id, kind, title, topic, slug, status, publication/gate settings, timestamps, available_post_variants, editor_url, and public_url; plus count.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only; no write approval.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-list_projects",
  "method": "tools/call",
  "params": {
    "name": "list_projects",
    "arguments": {
      "kind": "post",
      "limit": 25
    }
  }
}

Read one owned Guide's title, publication and lead-gate settings, current questions, and reusable questions from other Guides.

How it works

Reads one owned Guide's editable public settings and normalized lead questions. It also deduplicates and returns reusable questions from the account's other Guides.

Returns

project_id, title, status, published, noindex, email_gate, gate_questions, reusable_questions, public_url, editor_url, and leads_url.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only. Call this before update_guide_settings or set_guide_questions.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-get_guide_settings",
  "method": "tools/call",
  "params": {
    "name": "get_guide_settings",
    "arguments": {
      "project_id": 456
    }
  }
}

Update an owned Guide's title, public/private state, lead gate, and search-engine visibility. Publishing requires completed Guide content.

How it works

Patches only the supplied Guide settings. title changes the displayed title; published controls the public URL; email_gate controls the name/email gate; noindex controls search-engine indexing. Publishing is rejected until the Guide contains completed Markdown.

Returns

The same complete settings payload as get_guide_settings, plus updated_fields and credits_charged: 0.

Credits

0 credits; no LLM call.

Approval / side effect

Account/public write. Publishing or unpublishing changes external visibility and must be shown for approval.

• At least one setting besides project_id is required.
• Omitted settings remain unchanged.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-update_guide_settings",
  "method": "tools/call",
  "params": {
    "name": "update_guide_settings",
    "arguments": {
      "project_id": 456,
      "title": "AI Agent Evals: The Complete Guide",
      "published": true,
      "email_gate": true,
      "noindex": true
    }
  }
}

Replace the complete lead-question list for one owned Guide. Read current settings first and include existing questions you want to keep. Supports text, single-answer, and multiple-answer questions.

How it works

Replaces the complete lead-question list for one owned Guide and optionally changes the name/email gate. Question IDs are normalized when missing or duplicated. single and multiple questions require 2-12 distinct choices; text questions store no choices.

Returns

The complete saved Guide settings payload, normalized gate_questions, reusable_questions, and credits_charged: 0.

Credits

0 credits; no LLM call.

Approval / side effect

Account write. Read current questions first and show the complete replacement list before saving.

• An empty questions array removes every custom question.
• Maximum: eight questions per Guide.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-set_guide_questions",
  "method": "tools/call",
  "params": {
    "name": "set_guide_questions",
    "arguments": {
      "project_id": 456,
      "email_gate": true,
      "questions": [
        {
          "id": "challenge",
          "prompt": "What is your biggest agent-evaluation challenge?",
          "type": "text",
          "required": true
        },
        {
          "id": "stage",
          "prompt": "What stage are you at?",
          "type": "single",
          "required": true,
          "options": [
            "Planning",
            "Pilot",
            "Production"
          ]
        }
      ]
    }
  }
}

Create a free 1080×1080 PNG from an owned Post draft or Guide. Returns the image directly and an expiring download link.

How it works

Renders a deterministic 1080x1080 PNG from an owned Guide or selected Post draft. word_cloud extracts prominent terms; quote_card uses the selected content's opening hook. The MCP result contains the image itself and a signed download resource.

Returns

Structured metadata (project_id, content_source, style, filename, MIME type, dimensions, digest, download_url, expiry, credits_charged) plus an MCP image content block and resource_link.

Credits

0 credits; local rendering, no LLM call.

Approval / side effect

Creates a derived file but does not publish or upload it.

• variant is required for a Post project; omit it to use Guide Markdown.
• The signed download URL expires after 3,600 seconds.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-create_social_image",
  "method": "tools/call",
  "params": {
    "name": "create_social_image",
    "arguments": {
      "project_id": 789,
      "variant": 2,
      "style": "word_cloud"
    }
  }
}

Create a free expiring download link for a styled 1080×1080 five-second MP4 from an owned Post draft or Guide. Rendering starts when the link is downloaded.

How it works

Creates a signed resource link for a styled 1080x1080, five-second scrolling MP4 based on an owned Guide or Post draft. Rendering begins only when the signed URL is opened, so the tool call itself returns quickly.

Returns

project_id, content_source, filename, MIME type, 1080x1080 dimensions, five-second duration, download_url, expiry, rendering note, credits_charged, and an MCP resource_link.

Credits

0 credits; local video rendering, no LLM call.

Approval / side effect

Creates a private derived download; it does not upload or publish media.

• variant is required for a Post project.
• The link expires after one hour and permits at most six video downloads per account per hour.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-create_scroll_video",
  "method": "tools/call",
  "params": {
    "name": "create_scroll_video",
    "arguments": {
      "project_id": 789,
      "variant": 2
    }
  }
}

List the authenticated creator's guide projects and recipient-link open totals.

How it works

Lists owned Guide projects newest first. published optionally filters to public or private Guides; omit it to list both.

Returns

guides with id, title, slug, status, publication/noindex state, updated_at, tracked-link/open totals, and public_url; plus count.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only; no write approval.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-list_guides",
  "method": "tools/call",
  "params": {
    "name": "list_guides",
    "arguments": {
      "published": true,
      "limit": 25
    }
  }
}

Read one owned guide, including its Markdown content and publication URL.

How it works

Reads one owned Guide, including its full Markdown and public settings. It does not increment public Guide analytics.

Returns

id, title, topic, slug, status, published, noindex, email_gate, guide_markdown, timestamps, and public_url.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only; no write approval.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-get_guide",
  "method": "tools/call",
  "params": {
    "name": "get_guide",
    "arguments": {
      "project_id": 456
    }
  }
}

Create a unique public-guide link for one LinkedIn recipient. Opens are confirmed only after a real guide page loads.

How it works

Creates one unique /r/ recipient URL for an owned, published Guide. The optional LinkedIn URL must be a public linkedin.com profile. The result also includes a short draft message. GuideBlink records opens but never sends the message.

Returns

Link identity, project/user fields, recipient name, LinkedIn URL, private note, active state, timestamps, unique public url, and draft message. The internal token is not returned separately.

Credits

0 credits; no LLM call.

Approval / side effect

Creates an external-facing unique URL. The user must review and send any DM separately.

• The Guide must already be published.
• The account limit is 10,000 recipient links.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-create_tracked_link",
  "method": "tools/call",
  "params": {
    "name": "create_tracked_link",
    "arguments": {
      "project_id": 456,
      "recipient_name": "Morgan Lee",
      "linkedin_url": "https://www.linkedin.com/in/morgan-lee",
      "note": "Asked about agent evals"
    }
  }
}

List recipient-specific links with confirmed open counts and first/last-opened timestamps.

How it works

Lists recipient links newest first, optionally scoped to one owned Guide. Counts represent confirmed page loads after preview-bot filtering; they do not prove the named person opened the link.

Returns

links with recipient/project data, unique URL, draft message, active state, confirmed open and unique-opener counts, first/last-open timestamps, response_count; plus count.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only; no write approval.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-list_tracked_links",
  "method": "tools/call",
  "params": {
    "name": "list_tracked_links",
    "arguments": {
      "project_id": 456,
      "limit": 50
    }
  }
}

Read one owned recipient link, confirmed opens, and voluntary guide-form answers attributed to that link.

How it works

Reads one owned recipient link, its confirmed open summary, and up to 50 latest voluntarily submitted lead responses attributed to that link. Each response includes name, email, consent, timestamp, and normalized question answers.

Returns

The complete tracked-link summary, responses, total response_count, responses_returned, and attribution_note.

Credits

0 credits; no LLM call.

Approval / side effect

Read-only, but results contain personal lead data and should be handled privately.

• A forwarded unique link can be opened or answered by someone other than the named recipient.

Parameters

{
  "jsonrpc": "2.0",
  "id": "call-get_tracked_link_activity",
  "method": "tools/call",
  "params": {
    "name": "get_tracked_link_activity",
    "arguments": {
      "link_id": 987
    }
  }
}