> 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/platform/slack-and-teams.md). # Where Agents Run Where agents run: the agent-first MCP surface, team chat, and the Pippa control room on one gateway and one audit trail. > **Same gateway, same governance, same audit trail — wherever the work happens.** Agent-first integration (your harness, IDE, or pipeline over MCP) is the primary surface. Team chat and the Pippa control room are the two human-native surfaces on the same platform. ## Surfaces at a glance | Surface | Who uses it | What it is for | | --------------------- | --------------------------- | ------------------------------------------------------------------------------- | | **Agent-first (MCP)** | Agents, harnesses, IDEs, CI | Primary surface — see [MCP Access](/connect/mcp-access.md) | | **Slack & Teams** | Teams working in a channel | Team-native conversational operations (this page) | | **Pippa Chat** | Individual operators | A web/mobile control room to run, watch, and direct media work conversationally | ## Pippa Chat Pippa is the control-room interface, not the front door: a consolidated web/mobile UI where a single operator can ask the platform to perform a media task in natural language, watch the run unfold (plan, tools called, results, costs), approve, override, or correct the agent mid-run, and drill from a task into the underlying Evidence Cards and trace-correlated audit rows. It uses the same gateway, governance, and audit chain as every other surface. ## Slack & Teams When the work happens in a team channel, Mediafier shows up there. *** ## Why a chat surface Most media operations are not solo. They are coordinated in shared team channels... content review threads, ingest queues, brand-safety escalations, distribution sign-offs. The work people already do in Slack and Teams is the work that benefits most from a governed AI agent reaching it directly. The Slack & Teams surface is purpose-built for that flow. An agent joins the channel; the team converses with it the way they converse with each other; every action the agent takes is identity-bound, metered, and audited. *** ## Where chat sits in the surface order The three primary surfaces, in this order: 1. **Agent-first (CLI + MCP)** — primary integration. 2. **Slack & Teams** — team-native conversational operations. 3. **Pippa chat** — individual control-room interface. Chat surfaces are second-class to agent-first because they are a different audience: agent-first is for the engineers integrating Mediafier; chat surfaces are for the teams running on Mediafier. *** ## What chat is good at | Use case | Why chat | | ---------------------------------------- | ------------------------------------------------------------------------------ | | Approvals routed to the right humans | The approval lands in the channel where the relevant operators already are. | | Notification + drill-down | The agent posts a status, with a thread for follow-up. | | Multi-operator collaboration | Two operators can react to the same run inside the same thread. | | Triage from existing escalation channels | The agent enters the channel where escalations already happen, not a new tool. | *** ## What chat shares with every other surface The same gateway runs every chat-originated tool call. That means: * **Identity** — the chat user's identity is mapped server-side to a Mediafier principal via Auth0. The agent never trusts a client- supplied identity claim. * **Org isolation** — the tenant context is derived from the resolved principal. A user in the wrong org cannot reach another org's tools via chat. * **Authorization** — only tools the chat user is permitted to invoke are reachable. * **Rate limiting** — chat-originated calls share the org's rate budget. * **Billing** — every tool call is metered against the org's credit ledger. * **Audit** — every action produces an immutable audit row and a trace ID. A run kicked off from a Slack thread is indistinguishable from a run kicked off from a CLI as far as the audit chain is concerned. Chat is presentation; the chain underneath is the same chain. *** ## What chat is not * It is not a separate identity domain. The chat user is mapped to a real Mediafier principal; chat does not get a fast-path bypass. * It is not a privileged client. Same gateway, same audit, same credits. * It is not the only place to operate the platform. Agents can be driven from any harness; individual operators can use Pippa. *** ## Where to go next | If you're… | Read next | | ----------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------ | | Engineering an integration | [Agent-First Access (CLI + MCP)](/connect/agent-first-access.md) | | Looking for the individual control-room surface | Pippa Chat (above) | | Reviewing identity and tenancy for chat users | [Identity & Authentication](/trust/identity-and-authentication.md), [Multi-Tenant Organization Model](/trust/multi-tenant-organization-model.md) | | Reviewing audit posture | [Audit, Observability & Compliance](/trust/audit-observability-and-compliance.md) |