# Pixero > Pixero is an AI marketing agent that handles content, influencers, and ads. Over MCP, given a website URL and a budget, it researches the brand, writes the strategy, generates the ad images and video, and launches and manages the Meta campaign. A human approves before anything spends money. ## What it is Most ad-related MCP servers are read/write wrappers over an ad platform's marketing API: they can report on campaigns you already have and change budgets on ads someone else made. Pixero also makes the ads. Creative generation, campaign structure, launch, and optimization are one loop behind one connection, and the same loop is available to an assistant over MCP, to a developer over HTTP, and to a platform embedding it for its own customers. ## How to connect - Remote MCP endpoint: https://pixero.ai/api/mcp - Transport: streamable HTTP - Auth: OAuth (any MCP client walks you through it), or a Pixero API key (`pxr_...`) passed as a bearer token - Official registry entry: `io.github.ryanzambrano/pixero` - Setup guide for Claude and other MCP clients: https://pixero.ai/docs/claude - API keys and scopes: https://pixero.ai/docs/api-keys ## Permission model Scopes are checked per call, not at connect time, so a key without `deploy` connects successfully and then fails at the launch step. - `read`: reporting, research, and reading your own data. Costs nothing and is never behind a subscription check. - `generate`: image and video generation. Metered in credits. - `deploy`: anything that publishes to an ad account or configures an integration. - `agent`: delegates a whole task to the Pixero agent with its own toolset. ## What an agent can do end to end 1. Point it at a website. It researches the brand and builds a durable brand memory: identity, voice, audience, products, competitors. 2. Ask for research. It pulls a competitor's live Meta ads from the Ad Library and analyzes what those brands keep paying for. 3. Ask for creative. It generates ad images and video from the brand memory, metered in credits, with cost estimable before the call. 4. Ask it to launch. It builds the campaign, ad sets, and ads on the connected Meta account and publishes them paused, so a human activates delivery. 5. Ask how it went. It reads back performance per ad and can be told to scale, pause, or iterate on a winner. ## For platforms embedding Pixero A platform can offer ad launching to its own customers without building any of it. Guided integration setup binds an API key to an integration, each end customer is scoped by a `vendorRef` the platform already knows them by, launches are idempotent, and leads are delivered to a webhook with a signature. See https://pixero.ai/docs/platform. ## Tool reference ### Social publishing Use the same Outstand connections as the Publish page. - list_publishing_accounts (read): Find connected accounts. - list_publishable_media (read): Browse completed media. - create_post (deploy): Publish now or schedule an approved post. - list_posts (read): Read the queue and recent history. - get_post_status (read): Check per-account delivery outcomes. - cancel_post (deploy): Cancel an approved queued post. ### Creators Saved creators, notes, briefs, assignments and owner-authorized X messages. - list_creators (read): List creators. - get_creator (read): Get creator. - save_creator_notes (deploy): Save creator notes. - list_creator_briefs (read): List creator briefs. - get_creator_brief (read): Get creator brief. - save_creator_brief (deploy): Save creator brief. - assign_creator_brief (deploy): Assign creator brief. - list_creator_assignments (read): List creator assignments. - update_creator_assignment (deploy): Update creator assignment. - list_creator_conversations (read): List creator conversations. - get_creator_messages (read): Get creator messages. - send_creator_message (deploy): Send creator message. ### Setting up a platform integration Start here if you are embedding Pixero in your own product. These cover onboarding and the status your users see. - start_integration_setup (deploy): Begins guided setup and returns the first question to ask you: which organization this is for, then your business. One question at a time, resumable, and it ends by reading the resulting policy back before anything is saved. - answer_integration_setup (deploy): Records your answer to the current question and returns the next one. After the last, returns the summary and the launch policy derived from it. - submit_integration_setup (deploy): Saves the reviewed answers: creates the organization if you asked for a new one, applies the policy to it, and binds the API key to the integration. Only after you confirm. - get_vendor_status (read): One customer's campaign and leads, scoped to the vendorRef your platform knows them by, so each user sees only their own. ### Workspaces and brands Who the work belongs to. A brand workspace holds one brand profile, its campaigns, and its leads. - list_workspaces (read): The workspaces you can act in: every brand or agency organization you belong to, plus the client brands your agency manages. - create_brand_workspace (deploy): Create a customer's workspace under your agency, or return the existing one. Idempotent by name. - list_brands (read): The brand profiles you can access, optionally scoped to one workspace. - get_brand_memory (read): One brand profile in full, the same one the agent reads. - get_brand_competitors (read): The competitors a brand's research found, with their positioning and ads. - remember (deploy): Save a win, failure, correction or preference to the brand's memory for every later run. - analyze_brand_website (generate): Read a website and return draft onboarding answers to correct and save. - submit_brand_profile (deploy): Save those answers as the brand profile and start the research pipeline. - get_onboarding_status (read): Poll the research pipeline started by submit_brand_profile. - set_workspace_meta_defaults (deploy): Set the ad account, Page, and pixel a workspace launches with. Set on an agency and clients inherit. - get_workspace_integration (read): Read which platform integration a workspace is bound to, and its policy. - set_workspace_integration (deploy): Bind a workspace to a platform integration and its guardrails. ### Launching on Meta Everything that touches real ad delivery. All of it requires an explicit confirmation. - list_ad_accounts (read): Whether Meta is connected, which ad account and Page ids are valid, and what is missing if not. Call this first. - autopilot_launch_campaign (deploy): A brief to a live campaign in one call: strategy, creative, review, publish, activate. - meta_launch_action (deploy): Act on a staged plan: status, publish, activate, pause, cancel a stuck publish, or unwind a wrong one. - upload_creative_to_meta (deploy): Put an image into a Meta ad account's creative library. - get_ad_performance (read): Synced Meta performance, sorted by ROAS, spend, purchases, or CPA. ### Leads On-Facebook lead capture and delivery. Website conversions do not pass through here. - list_lead_forms (read): Instant Forms on a Page, or the Pages themselves when no id is given. - create_lead_form (deploy): Create an Instant Form on a Page and subscribe it to lead webhooks. - set_lead_destination (deploy): Route a workspace's leads to your endpoint and return the signing secret. - get_lead_destination (read): Read where a workspace's leads currently go. - list_leads (read): Captured leads and their delivery status, including failures. ### Creative Generation spends credits. Estimate first if the cost matters. - generate_image (generate): Create an image, optionally from reference images. - generate_video (generate): Create a video from a scene prompt, with optional start and end frames. - get_video_status (read): Poll a video generation started earlier. - list_video_models (read): Available video models with their durations and aspect ratios. - estimate_video_cost (read): Credit cost of a model and duration before you commit. - get_credit_balance (read): Credits available on the account behind the key. ### Research and delegation Open-ended work, and the competitor research behind it. - search_competitor_ads (read): A brand's live Meta ads with per-ad analysis and a whole-brand summary. - list_canvases (read): Briefs across every workspace you can reach. - run_pixero_task (agent): Delegate a task to the Pixero agent with its full toolset. Returns handles to poll. - get_task_status (read): Check a delegated task and read its final message. ## Pages ## Product - [Pixero](https://pixero.ai): Overview of the agent: give it a website URL and a budget, it researches the brand, generates ad images and video, and launches and manages Meta campaigns with a human approving each step. - [Pricing](https://pixero.ai/pricing): Subscription tiers and the credit system that meters image and video generation. Reading your own data over the API costs nothing; credits gate generation and publishing gates spend. - [Portfolio](https://pixero.ai/portfolio): Ad creative Pixero generated, shown as produced. Demo work comes from running the product on a fictional brand rather than from mockups. - [Managed service](https://pixero.ai/managed): High-touch engagement where the Pixero team runs the account, sold through a call rather than self-serve signup. ## MCP and API - [Pixero MCP server (ad automation for agents)](https://pixero.ai/mcp): The remote MCP server at https://pixero.ai/api/mcp: what the tools do, how OAuth and API-key auth work, and how an agent goes from a brand URL to a live Meta campaign without a human opening Ads Manager. - [Documentation](https://pixero.ai/docs): Index of the developer documentation: connecting an assistant, API keys and scopes, the full tool reference, platform embedding, Ad Spy, and the conversion pixel. - [Docs: overview](https://pixero.ai/docs/overview): How the Pixero API is put together: workspaces, brands, credits, scopes, and the difference between reading data and spending money. - [Docs: connect Pixero to Claude](https://pixero.ai/docs/claude): Adding the Pixero MCP connector to Claude and other MCP clients over OAuth, and what the assistant can do once connected. - [Docs: MCP tool reference](https://pixero.ai/docs/tools): Every tool the MCP server exposes, grouped by the job it does, with the permission scope each one checks. - [Docs: API keys and scopes](https://pixero.ai/docs/api-keys): Creating a pxr_ API key, the read, generate, deploy, and agent scopes, and using a key as a bearer token against the MCP endpoint. - [Docs: embed Pixero in your platform](https://pixero.ai/docs/platform): The vendor path: guided integration setup, per-customer vendorRef scoping, embedded launch widgets, and signed webhooks for lead delivery. - [Docs: Ad Spy API](https://pixero.ai/docs/adspy): Pulling any brand's live Meta ads from the Ad Library through the Pixero API, for competitor research inside your own code or agent. - [Docs: conversion pixel](https://pixero.ai/docs/pixel): Installing the Pixero pixel so conversions attribute back to the ads the agent launched. - [Docs: Meta account setup](https://pixero.ai/docs/onboarding): Setting up Meta Business Manager and granting Pixero access to an ad account. - [Connect Pixero to Claude and use the Ad Spy API](https://pixero.ai/resources/mcp-api): Two-minute MCP connector setup for Claude, then one curl command to pull a brand's live Meta ads with an API key. - [Meta ads MCP servers compared](https://pixero.ai/articles/meta-ads-mcp-servers): Meta's official Ads MCP server at mcp.facebook.com/ads against Pipeboard, Foreplay, Ryze and Pixero, on write capability, permission scoping, network coverage and the absence of triggers in the protocol. ## Guides - [Resources](https://pixero.ai/resources): Index of setup guides and fundamentals, written for people running paid social for the first time. - [Meta ads fundamentals](https://pixero.ai/resources/meta-ads-fundamentals): Business account setup, pixel install, messaging, launching with CBO, and deciding which ads to keep. - [How to use Pixero](https://pixero.ai/how-to): End-to-end user guide: brand memory, directing the agent on the canvas, generating creative, Ad Spy research, and the staged Meta launch flow. ## Articles - [Articles](https://pixero.ai/articles): Index of the paid-social strategy playbooks. - [How to start running ads](https://pixero.ai/articles/start-running-ads): What to do in the first two weeks of a new ad account, and the mistakes that waste the first $1,000. - [Ad creative that converts](https://pixero.ai/articles/ad-creative-that-converts): What separates a creative that scales from one that dies in testing, hook by hook. - [How to clone competitor ads](https://pixero.ai/articles/clone-competitor-ads): Mining the Meta Ad Library for a competitor's proven winners and positioning against the gaps. - [How to scale winning ads](https://pixero.ai/articles/scale-winning-ads): Increasing spend on a winner without resetting learning or collapsing the return. - [Three things top marketers do](https://pixero.ai/articles/three-things-top-marketers-do): The habits that separate media buyers who compound from ones who restart every quarter. - [How to automate ads with AI](https://pixero.ai/articles/automate-ads-with-ai): Which parts of a paid-social workflow can be handed to an agent today, which still need a human, and where the handoff belongs. - [Best AI ad tools for a Shopify store](https://pixero.ai/articles/best-ai-ad-tools-for-shopify): AI ad tools for ecommerce sorted by the bottleneck they remove: static volume, video format, creative angle, attribution, or nobody making the ads at all. - [AdCreative.ai alternatives](https://pixero.ai/articles/adcreative-ai-alternatives): The four reasons people leave AdCreative.ai and the different product each one points to, with Cropink, Creatify, Madgicx and Pixero compared on what happens after the file downloads. - [Foreplay alternatives](https://pixero.ai/articles/foreplay-alternatives): Competitor ad research tools compared on workflow rather than data, since all of them resell Meta's public Ad Library, including which ones an AI agent can read directly. - [Arcads vs Creatify](https://pixero.ai/articles/arcads-vs-creatify): Two AI UGC ad generators compared on actor realism, published pricing and workflow, plus the creative strategy step neither of them covers. - [Madgicx review](https://pixero.ai/articles/madgicx-review): What Madgicx audits well on an established Meta account, how its seven ad spend pricing bands work, and the accounts that have nothing for it to analyze. - [Ad software for founders with no marketing team](https://pixero.ai/articles/ad-software-for-solo-founders): What one person can sustainably run in paid social, what to buy at each level of ad spend, and how the tooling cost compares with an agency retainer. - [How to export your ad data](https://pixero.ai/articles/export-meta-ads-data): Getting performance history and creative out of Meta Ads Manager, and which dashboards, rules, swipe boards and brand kits do not survive cancelling a third-party tool. - [Buffer vs Later vs Hootsuite vs Metricool](https://pixero.ai/articles/social-scheduling-tools-compared): Four social schedulers compared on published pricing, platform coverage and how each counts connected accounts, plus the content supply problem none of them addresses. - [Influencer platform alternatives](https://pixero.ai/articles/influencer-platform-alternatives): What influencer platforms charge for, why creator discovery is the commoditized half, and which parts of running a creator program the category leaves to the buyer. - [Turning creator content into paid ads](https://pixero.ai/articles/creator-content-into-paid-ads): How TikTok Spark Ads and Meta partnership ads work, the authorisation flow on each platform, and the usage rights and raw footage terms that belong in a creator brief. - [Ryze vs Madgicx vs Smartly vs Pixero](https://pixero.ai/articles/ryze-vs-madgicx-vs-smartly-vs-pixero): Four AI ad platforms compared on creative generation, campaign launch, channel coverage, agent access and price, with a section naming where Pixero loses to each of the other three. ## Company - [Agnost case study](https://pixero.ai/portfolio/case-studies/agnost): Pixero helped Agnost get 700k views on their launch video by working with content creators on Twitter. - [FAQ](https://pixero.ai/faq): Common questions about credits, video models, Meta permissions, billing, and what the agent will and will not do without approval. - [Talk to us](https://pixero.ai/lets-talk): Book a call about the managed service or a platform integration. - [Feedback](https://pixero.ai/feedback): Send a bug report or feature request. - [Privacy policy](https://pixero.ai/privacy): How Pixero handles customer and end-user data. - [Terms of service](https://pixero.ai/terms): Terms covering the product, credits, and ad spend.