# Convia for AI agents > The two ways AI agents work with Convia, reading these docs and acting on a workspace through MCP, plus what an agent can and cannot do and where to start. > For the complete documentation index, see [llms.txt](https://docs.conviapro.com/llms.txt). Convia Pro is built to work with AI agents as well as people. An agent can learn how Convia works by reading these docs, and it can act on a workspace, reading content and writing drafts, through the Convia MCP server. This page explains both, what an agent is and is not allowed to do, and where to start. ## Two ways an agent works with Convia | | Read the docs | Act on a workspace | |---|---|---| | What it is | Every docs page as plain markdown, plus an index and a single bulk file | A Model Context Protocol (MCP) server scoped to one workspace | | What the agent gets | How features work, settings, limits, how-tos | The workspace's own episodes, transcripts, clips, posts, ideas, campaigns and analytics, and the ability to write drafts | | Needs an account | No | Yes: an Owner or Manager, on a plan that includes MCP Connections | | Start here | [Machine-readable docs](https://docs.conviapro.com/docs/machine-readable-docs.md) | [Connect an AI assistant (MCP)](https://docs.conviapro.com/docs/mcp-quickstart.md) | Most agents use both: the docs to understand what a feature does, and MCP to do the work in a specific workspace. ## What an agent can do through MCP The Convia MCP server exposes 38 tools. With them an agent can: - **Learn the brand.** Read the workspace's [Brand Brain](https://docs.conviapro.com/docs/brand-brain.md) (voice, audience, content rules, host voice), its prompt templates, its reusable text snippets and its posting schedule. - **Find source material.** List episodes and guests, read transcripts with speakers and timestamps, search episodes, clips, posts and transcripts, and see which clips have been used least. - **Check claims.** Search transcripts for a figure or quote before writing it, so content rests on what was actually said. - **See what exists.** List posts, the calendar, ideas, scripts and campaigns, and see everything already made from an episode, clip or idea. - **Write drafts.** Create posts (with a clip or existing media attached), whole multi-platform campaigns, ideas, scripts and episode records. - **Revise its own work.** Read back and update posts, ideas, scripts and a campaign plan that has no posts yet. - **Take posts back.** Unschedule, archive or move posts to Trash, up to 25 at a time, always with a stated reason. - **Read results.** Post analytics, website traffic from Convia-tagged links, and the searches that bring people to the workspace's site (these need Advanced Analytics). - **Stage prompt changes.** Save a new version of a content prompt template for a manager to make live. The full list, with parameters and limits, is in the [MCP tool reference](https://docs.conviapro.com/docs/mcp-tools.md). ## What an agent cannot do Convia keeps decisions that publish, approve or declare something with a person. - **Nothing is published through MCP.** Every post an agent creates is a draft. A draft with a date sits on the calendar until a person schedules it, and a campaign stays a draft until a person activates it. Editing a post a person already scheduled leaves it scheduled. No tool approves a post or sends one to the [approval queue](https://docs.conviapro.com/docs/approval-and-review.md). - **Published and approved posts are off limits.** An agent cannot edit, archive or delete them. An edit to a post waiting for approval sends it back to draft, so the change is reviewed. - **Some publish settings stay with you.** An agent can set an Instagram post's format, whether a reel shows in the feed, and trial reels, plus YouTube category, playlists, tags and audio language. It cannot set TikTok privacy or branded-content disclosures, or YouTube's Made for Kids, privacy status or publish time. - **No new media.** An agent can attach clips and media already in the workspace, but cannot upload files. - **Live AI prompts are protected.** An agent cannot change the default prompt template or a per-platform social prompt. It can only stage an alternative. - **No deletions beyond posts.** Ideas are archived rather than deleted, and there are no delete tools for scripts, episodes or campaigns. - **Read-only settings.** Brand Brain, text snippets and posting schedule settings can be read, not changed. - **One workspace at a time.** A connection reaches only the workspace in its URL. ## Who can connect - **Role.** Only a workspace **Owner** or **Manager** can create a token or connect a client, and the role is checked on every request. - **Plan.** The workspace's plan must include MCP Connections. Post analytics and website analytics tools also need Advanced Analytics, and the campaign plan tools need [Magic Campaign](https://docs.conviapro.com/docs/magic-campaign.md). See [Plans & billing](https://docs.conviapro.com/docs/plans-and-billing.md). - **Clients.** Claude on the web, iPhone and Android and Claude Desktop connect with a custom connector that signs in to Convia. Claude Code, a Claude Desktop config file and most other MCP clients use a workspace token. Details are in [MCP access, permissions & audit](https://docs.conviapro.com/docs/mcp-security.md). ## Credits and accountability Using MCP spends no [AI credits](https://docs.conviapro.com/docs/ai-credits.md). The agent writes the copy on your own subscription with that AI assistant, and Convia stores the result. Every tool call is recorded on the **Activity** tab of **Integrations → MCP Connections** (under **Workspace** in the launcher), with the tool, the result and the time it took. Tokens can be revoked at any time, and removing a member or changing their role stops their access on the next request. ## Where to start 1. Read [Connect an AI assistant (MCP)](https://docs.conviapro.com/docs/mcp-quickstart.md) to connect Claude or another client. 2. Try the recipes in [Agent workflows](https://docs.conviapro.com/docs/mcp-workflows.md), starting with drafting one post. 3. Keep the [MCP tool reference](https://docs.conviapro.com/docs/mcp-tools.md) and the [Glossary](https://docs.conviapro.com/docs/glossary.md) at hand for exact names and terms. 4. Point your agent at [Machine-readable docs](https://docs.conviapro.com/docs/machine-readable-docs.md) so it can look up how any feature works. ## Related - [Connect an AI assistant (MCP)](https://docs.conviapro.com/docs/mcp-quickstart.md): set up a connection in a few minutes. - [Agent workflows](https://docs.conviapro.com/docs/mcp-workflows.md): tool sequences for common jobs. - [MCP access, permissions & audit](https://docs.conviapro.com/docs/mcp-security.md): roles, plan, tokens and the Activity log. - [Machine-readable docs](https://docs.conviapro.com/docs/machine-readable-docs.md): read these docs as markdown. --- Source: https://docs.conviapro.com/docs/ai-agents-overview · Section: AI agents & MCP · Last updated: 2026-10-03