Skip to main content
TryPost provides a built-in MCP (Model Context Protocol) server that lets AI assistants manage your social media scheduling directly. Your assistant can create, schedule, queue and publish posts, handle approvals, organize ideas, read analytics, and manage signatures, labels, posting times, webhooks, repurposes and API keys — all through natural language.

What is MCP?

MCP is an open protocol that connects AI assistants to external tools and data sources. Instead of copying and pasting API calls, your AI assistant can interact with TryPost directly.

Available tools

The server exposes 83 tools. See the Tools reference for every argument and the recommended workflow.

What MCP cannot do

  • Generate content with AI. There is no MCP tool for generating or rewriting content. Your assistant writes the text itself and sends it to TryPost; the built-in AI writing assistant is available in the web app. Repurpose may shorten captions with AI in the background, including for rules created through MCP.
  • Connect, reconnect or disconnect channels, turn ideas into posts, manage templates, members, invites or billing. Use the web app for these.

How it works

MCP on TryPost Cloud needs an account with access to the app (an active subscription). See Plans & billing. Self-hosted instances skip this check.
1

Add the TryPost MCP server in your assistant

Use your assistant’s MCP configuration to add a remote HTTP server pointing to https://app.trypost.it/mcp/trypost. Each client guide below has the exact command or settings file.
2

Sign in and pick a workspace

The first time the assistant connects, TryPost opens a browser window. Sign in, choose the Workspace this connection may use, and click Authorize. The assistant holds the OAuth token from then on, with automatic refresh.
3

Start using natural language

Ask your assistant to manage your TryPost account. For example: “Queue a LinkedIn post for my next free slot” or “List my posts waiting for approval.”

Server URL

Self-hosting TryPost? The server lives at {APP_URL}/mcp/trypost on your instance.

Authentication

TryPost’s MCP server uses OAuth 2.1 with Dynamic Client Registration (DCR), PKCE, and the mcp:use scope. Modern MCP clients detect this automatically from the .well-known discovery endpoints — you only point them at the URL and the browser sign-in handles the rest. Each connection is bound to the workspace you chose on the consent screen (“This connection will only access the selected workspace.”). To work in another workspace, connect again and pick it. Manage or disconnect connected assistants under workspace Settings → MCP.
Personal Access Tokens from Settings → API Keys authenticate the REST API only. They are not accepted on /mcp/trypost. MCP requires an OAuth grant with the mcp:use scope (browser sign-in from a supported client).

Permissions

Every member of the workspace can connect an assistant, not only admins (the REST API accepts admin API keys only). The assistant can do what you can do in the app:
  • Any member can write posts, notes, ideas, labels and signatures and read analytics.
  • Members who need approval: posts they schedule, queue or publish wait as pending_approval until a member who publishes directly approves them. Ask the assistant to call get-workspace-tool to see which applies to you.
  • Members who publish directly can also approve and reject posts, manage the queue order, set recurrence and run repurposes.
  • Admins can also change posting times and manage webhooks and API keys.
See Members and permissions and Approvals.

Troubleshooting

Your OAuth grant is valid and the config is correct — the account behind it just doesn’t have an active subscription. Open app.trypost.it, pick a plan, then reconnect the MCP server. Self-hosted instances skip this check.
The OAuth session expired and refresh failed, the access token is missing, malformed or revoked (removing you from the workspace revokes it), or the workspace it was bound to was deleted (No workspace selected.). Re-run the assistant’s connect flow to sign in again.
The request used a Personal Access Token (API key) or an OAuth token without the mcp:use scope. Remove any Authorization: Bearer … API key from the MCP config and connect with OAuth instead. For scripts and CI, call the REST API with an API key (workspace admins only) — not the MCP endpoint.
The connection is bound to a workspace you can no longer access. Reconnect and pick a workspace you belong to.
Run the connection probe your assistant exposes (for example claude mcp list or devin mcp list). The HTTP status from the probe tells you which case you’re in: 402 → see above; 401 → re-authenticate; 403 → see the two 403 entries above (use OAuth, not an API key; or reconnect to a workspace you belong to); anything else → open an issue on GitHub.
Some clients print the authorize URL to the terminal instead of launching the browser when running over SSH or inside a container. Copy the URL into a local browser, complete the flow, and the assistant will pick up the callback automatically. If the callback port is blocked, use a client that can finish OAuth locally, or drive TryPost through the REST API with an API key (workspace admins only).

Next steps

Every Build with AI page is linked here — the tools reference plus all 18 setup guides.

Tools reference

Every MCP tool, parameters, and the publish workflow.

Claude web

Claude on the web (claude.ai).

Claude Code

CLI assistant by Anthropic.

Claude Desktop

Desktop app by Anthropic.

ChatGPT

OpenAI ChatGPT with MCP apps.

Muse AI Agent

Personal agent by Meta.

Muse Code

CLI assistant by Meta.

Perplexity

Perplexity chat and Computer.

Grok

Grok on the web, iOS, and Android.

Cursor

AI-powered code editor.

Devin

AI software engineer by Cognition.

VS Code

Native MCP support in Copilot Chat.

GitHub Copilot

Copilot agent mode.

Gemini CLI

CLI assistant by Google.

OpenCode

Open-source CLI assistant.

OpenClaw

Personal AI assistant gateway.

Hermes

Hermes Agent by Nous Research.

Codex

CLI agent by OpenAI.

Antigravity

AI agent by Google.