Adros AI - Approval-Gated AI Marketing Workroom

Showing 25 of 79 from the live probe.

• Perplexity Research

  Run live Perplexity research with citation-rich output. Charges 4 Adros credits per request and uses the user's Perplexity BYOK secret when available, otherwise falls back to PERPLEXITY_API_KEY.

• Shopify List Products

  List Shopify products for catalog health checks before Shopping/PMax planning.

• Shopify Get Aov

  Compute Shopify average order value for ROAS forecasting.

• Shopify Get Channels

  Report Shopify granted scopes and channel-detection readiness.

• Search Patterns

  Search and filter the Adros pattern database. Returns lightweight pattern summaries sorted by quality or performance score. Each result includes image_url (the winning ad visual) so you can see what the ad looks like. Use this to discover which patterns exist for a given style or industry before calling get_blueprint for full details. Args: creative_style: e.g. "problem_solution", "us_vs_them", "negative_hook", "testimonial", "feature_benefit

• Get Blueprint

  Get the complete generation blueprint for a pattern. Returns full visual specs, text overlays, copy templates, psychology triggers, color strategy, layout rules, and generation instructions. Provide either pattern_id (UUID) or slug (e.g. "ryze-problem-solution-ad"). Use this after search_patterns or match_strategy to get actionable creative specs. Args: pattern_id: UUID of the pattern (from search_patterns or match_strategy results) slug: Slug of the patter

• Match Strategy

  Match winning ad patterns to a campaign strategy. The primary tool for the Adros pipeline. Given campaign inputs, returns Safe/Standard/Bold pattern recommendations with reasoning and industry benchmarks. Args: industry: Industry slug, e.g. "b2b_saas", "ecommerce", "health_wellness", "beauty", "fitness", "finance", "real_estate", "education", "food_beverage", "travel", "pet_care", "home_decor" awareness_stage: Where the target aud

• List Industries

  List all supported industries with their benchmarks. Returns industry slugs, names, default creative styles/frameworks, and performance benchmarks (CTR, CPC, CPM, ROAS). Use this to discover valid industry slugs for match_strategy.

• Get Taxonomy

  Get the full Adros creative taxonomy — all valid filter values for the pattern DB. Returns all valid values for: creative_styles, copy_frameworks, psychology_triggers, emotional_triggers, awareness_stages, alignment_categories. USE WHEN: - You need to know what creative_style values exist before calling search_patterns (e.g., is it "testimonial" or "social_proof"? is it "bold_claim" or "bold_statement"?) - You want to see all available copy frameworks to pick the righ

• Log Deployment

  Deprecated compatibility surface; no deployment state is written. Pattern reuse truth requires verified bounded-deploy lineage and an idempotent server-owned evidence key. Caller-supplied pattern/ad identifiers are not accepted as proof, so this tool remains discoverable for existing clients but always fails closed without opening a database session. Args: pattern_id: UUID of the pattern that was deployed platform: Ad platform: "meta" (default), "google", "t

• Save Business Context

  Save or update scoped business context for a user/client. Pass client_name when managing a brand; business_name is display text, not the scope key. Fill as many fields as possible because the Optimization Loop flags missing landing page, conversion source, brand colors, and stale context. ad_language must be a compact language label/code or comma-separated list no longer than 50 characters; put strategy prose in positioning and product/offer detail in product_description. raw_research is me

• Get Business Context

  Retrieve the stored business context for this user (or a specific client). Always call this at the start of any session to understand the business before taking any action. Returns the full business profile including positioning, competitors, goals, and what the brand can/cannot say. If the user manages multiple clients, pass client_name to load that client's context. If no client_name is provided, loads the default (solo) context. Paid users who manage multiple clients

• List Clients

  List all clients/brands this user manages. Returns a list of client names and IDs. Use this to see which clients have been onboarded. If the list is empty, the user is a solo operator and doesn't need to pass client_name to other tools. Args: user_token: Optional. Auto-injected from MCP connection URL. Leave empty.

• Get Adros Workspace Resume

  Use this as the first tool in every new external-AI session. It resumes durable brand state before choosing or rerunning an Adros stage. Do not begin or redo Research until this result confirms what already exists, what the user should review, and the next unfinished action. The inspection is read-only: it resolves an existing entitled brand, never creates a Project, never writes an event, and never contacts a paid provider. For a persisted brand it also includes a suppl

• Get Adros Workspace History

  Read one newest-first page of durable workspace history. The page unions MCP action receipts, agent session events, task approvals, and recorded outside actions for the authenticated workspace. Pass the returned opaque next_cursor to continue; a null next_cursor means history is exhausted. Actions never sent through MCP, a verified webhook or import, or reconciliation cannot be known automatically — record those with record_external_workspace_action. This inspection is

• Record External Workspace Action

  Record one action that happened OUTSIDE MCP so history stays honest. Use for manual console changes and imports (source_kind: manual, import). Verified webhook and rescan actions arrive only through server-side verified ingestion under a tenant-bound service identity, never through this tool; those source kinds are rejected typed before any read or write. Supply a stable external_event_id; redelivery replays the original receipt instead of duplicating it. Only a sanitize

• Get Commerce Readiness

  Read the durable catalog, Shopping, lead-form, and audience readiness contract. Resume the exact brand first, then call this before planning or attempting ecommerce/catalog work. This tool performs no provider call and cannot prove live readiness by itself. Use the returned setup blockers to run only the relevant read-only Google/Meta checks, then reconcile approved observations and read this exact brand again.

• Reconcile Commerce Readiness

  Persist approved read-only commerce observations for one exact brand. This writes only Adros's durable readiness contract. It never contacts or mutates Google, Meta, Shopify, or Merchant Center. Outside-AI observations are stored only as unverified attestations, even when the caller labels them provider readback. They require live read-only provider verification before readiness can become verified. This Adros-state write requires explicit user approval; provider mutatio

• Begin External Mcp Research Run

  Begin or resume one exact-brand external Research ownership graph. Use only after get_adros_workspace_resume confirms a fresh brand has no reusable Research package, the required intake is complete, and the user approves starting Research. Never use it merely because this is a new chat. This creates only the durable Project/Task lineage required to receive the canonical seven Research artifacts. It never launches an agent or calls a paid provider. The feature gate is che

• Save External Mcp Research Artifact

  Persist one artifact in the exact seven-file Research package. Call this only with lineage returned by begin_external_mcp_research_run. Use exactly research/<canonical filename> (or that exact bare canonical filename). For the five Markdown files, send raw UTF-8 in content_text. For the two .xlsx workbooks, prefer workbook_sheets so Adros renders the opaque XLSX bytes and computes their hash on the server. Binary-capable clients may still use content_base64, including sa

• Reconcile External Mcp Research Artifact

  Reconcile one stale or uncertain Research receipt without rewriting it. Use the client_name, project_id, run_id, and receipt_id returned by get_adros_workspace_resume. Use the receipt_id returned by get_adros_workspace_resume; never use an old chat's idempotency key. Do not resend artifact bytes: this tool reads bounded storage evidence and returns the safe next action without accepting raw ownership from the caller.

• Save Persona

  Save a customer persona discovered during market research. Call this for each distinct customer archetype identified. Personas inform creative angles, ad copy language, and audience targeting decisions. Args: user_token: Optional. Auto-injected from MCP connection URL. Leave empty. name: Persona name (e.g. "Busy Mom Sarah", "Tech-Savvy Tim") age_range: Age range (e.g. "25-34") pain_points: List of pain points this persona experiences desires:

• Get Personas

  Get all customer personas for this user (or a specific client). Use these to inform creative angles, ad copy, and targeting decisions. Each persona includes pain points, desires, objections, and the exact language they use — critical for writing ads that resonate. Args: user_token: Optional. Auto-injected from MCP connection URL. Leave empty. client_name: Optional — filter personas for a specific client/brand. Leave empty for solo users or t

• Log Weekly Action

  Log meaningful work for memory and optimization follow-up. Use after campaign changes, creative work, audits, deployment, or experiments. Include concise action_description, reason, before/after metrics when available, and client_name/client_id so future CEO and CMO sessions can explain what changed.

• Recall Weekly Log

  Recall the last N weeks of actions and their results. Always call this before running an audit so you know what was already tried, what worked, and what didn't. This prevents the AI from repeating failed strategies or missing successful patterns. Args: user_token: Optional. Auto-injected from MCP connection URL. Leave empty. weeks_back: How many weeks of history to retrieve (default 4) client_name: Optional — filter logs for a specific client/brand.