> For the complete documentation index, see [llms.txt](https://docs.mediafier.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.mediafier.ai/overview/what-is-mediafier.md). # What is Mediafier? The problem Mediafier solves, what it is (three layers, one enforcement chain), and who it is for. > **The agent-first control plane for media operations.** Content is becoming agent-addressable and every API is becoming an agentic endpoint. Mediafier is the governed layer that turns that shift into media work an organization can explain, repeat, and stand behind. *** ## The problem Generative and agentic AI is **probabilistic**. Legacy media pipelines are **deterministic**. Agents need rich operational context (provenance, lineage, edit history, brand-safety state, rights) to make decisions that hold up in production. Today, that context is spread across systems that weren't built for AI-native access, behind integration patterns that don't survive contact with autonomous agents — without shared identity, consistent authorization, per-call billing, or an audit trail an AI can hand a compliance officer. The result is a wide gap between what AI agents *could* do in media and what an enterprise can *actually* deploy. *** ## What Mediafier is Mediafier is **agent-first infrastructure for media** — three layers that work as one: 1. **A governed execution layer.** Connect the agent stack you already use — any MCP-aware agent, any harness, any IDE — to media systems and capabilities through one governed endpoint, without giving up control. Identity, permissions, spend, and audit are enforced consistently before any action runs. 2. **A media-native capability layer.** The Media Library, content intelligence, partner tools, and workflows become agent-addressable — the supply-chain, creation, and management primitives a media operator actually needs, exposed as governed tool calls. 3. **A compounding expert layer.** MediaClaws — domain super-agents — own repeatable media outcomes and learn from governed evidence and operational decisions, so repeated work improves over time. Operators stay in the loop where their judgment matters and step out of the loop where it doesn't. *** ## How customers reach it (in order) | Surface | What it's for | When to lead with it | | --------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- | | **Agent-first (CLI + MCP)** | Code-first, callable from any agent framework, IDE, or orchestrator. The platform shows up as native tools on whatever harness the team is already running. | Default for engineering integration, automation, CI, and any agent platform a team has already adopted. | | **Slack & Teams** | Team-native conversational operations. Agents join the channels operators already use. | Default for team workflows where the work happens inside a chat surface. | | **Pippa chat** | Web/mobile control-room interface. A consolidated view for monitoring, approvals, and direct conversational tasking. | Default for individual operators who prefer a single web/mobile UI for cross-tool work. | Surface order is intentional. Mediafier is agent-first; Pippa is one surface among several, not the platform's front door. *** ## Choose your path * **Media company** — evaluating or operating the platform: this page, then [Getting Access](/overview/access.md), [Getting Started](/connect/getting-started.md), and [Credits & Billing](/overview/credits-and-billing.md). * **Tool provider** — bringing your system to the marketplace: [The Marketplace](/partners/marketplace.md), then [Publishing to Mediafier's Marketplace](/partners/publishing-to-the-marketplace.md). * **AI agent** (or the human configuring one) — connecting right now: [Agent Quickstart](/connect/agent-quickstart.md), then [MCP Access](/connect/mcp-access.md) for harness-specific setup. *** ## Go deeper Each claim above has a page that earns it: | Pillar | The promise | The page | | -------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------- | | **The single gateway** | Every tool call from every surface crosses one enforcement chain — identity, organization, authorization, spend, audit — with a trace ID on every response. | [The Gateway](/platform/gateway.md) | | **Governed discovery** | Agents discover only what their organization may invoke, and bind to stable contracts that survive platform change. | [Governed Resources & Discovery](/platform/governed-resource-substrate.md) | | **MediaClaw super-agents** | Structured, bounded runs that leave an inspectable evidence trail and improve only under human-approved governance — never by fine-tuning on your data. | [MediaClaw Super-Agents](/platform/mediaclaw-super-agents.md) | | **Decision memory** | Every approval, override, escalation, or edit becomes searchable precedent that future runs reuse. | [Decision Tracing & Learning Loop](/platform/decision-tracing-and-learning-loop.md) | | **Media-native context** | Bundles, versions, MEDIA.md, and the Provenance Passport keep the rights, history, and intent attached to the work. | [Media Library](/platform/media-library.md) | *** ## What you can rely on | Guarantee | What it means in practice | | -------------------------- | ------------------------------------------------------------------------------------------------------------------ | | **Identity** | Auth0 is the sole identity provider. | | **Execution path** | All tool execution routes through the gateway — the only path to execution. | | **Tenant context** | Org context is derived server-side from your authenticated identity — never from a client header. | | **Discovery vs execution** | Discovery and planning are governed at the platform layer; execution remains with the runtime that ships the tool. | | **Audit correlation** | Every error includes a trace ID for support correlation. | | **Evidence immutability** | Evidence records are append-only and immutable — enforced at the platform's data layer, not by convention. | | **Learning governance** | Heuristics are extracted from runs; promotion to a live skill is human-approved. | | **Workspace scope** | Workspaces are run-scoped artifact, replay, and snapshot environments — not unrestricted shells. | *** ## Where to go next | If you're… | Read next | | ------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | | Sizing whether Mediafier fits your platform | [Core Philosophy](/overview/core-philosophy.md), [Platform Architecture](/platform/platform-architecture.md) | | Integrating from an agent harness or IDE | [Agent Quickstart](/connect/agent-quickstart.md); [MCP Access](/connect/mcp-access.md) | | Building or tuning a domain agent | [MediaClaw Super-Agents](/platform/mediaclaw-super-agents.md) | | Evaluating identity, audit, and tenancy | [Identity & Authentication](/trust/identity-and-authentication.md), [Audit, Observability & Compliance](/trust/audit-observability-and-compliance.md) | | Modeling spend and engagement | [Credits & Billing](/overview/credits-and-billing.md) |