Radi8 MCP connector reference | Radi8 Skip to main content
Sign In Try Radi8 free
Try Radi8 free

Connector reference

Radi8 MCP connector

Radi8 is a hosted remote MCP server. It holds a person's brand context, content strategy, monthly plan, drafts, schedule, and published posts, and it schedules posts to their connected channels (LinkedIn, X, Instagram, Facebook, TikTok). Radi8 is listed in Claude's connector directory and is an approved ChatGPT plugin; any client that supports remote MCP servers can connect with the details below.

Just want Radi8 in ChatGPT or Claude? The install guide gets you connected. Agents can read this page as markdown.

Endpoint https://app.radi8.com/api/mcp
Transport Streamable HTTP, stateless
Sign-in OAuth 2.1 with PKCE and dynamic client registration, or an API key
Scope mcp
Tools 36, in 9 groups (below)

Tested clients

ClientHow it connects
Claude claude.ai, Claude Desktop, and the Claude mobile apps. Add Radi8 from the connector directory. Listing
Install guide
ChatGPT The approved Radi8 plugin. Listing
Install guide
Claude Code Uses the connectors on the claude.ai account it signs in with, so Radi8 added in Claude appears on its own. Or add it to a project: claude mcp add --transport http radi8 https://app.radi8.com/api/mcp, start a new session, then run /mcp and choose Authenticate. Install guide

Any other client that supports remote MCP servers can connect with the details below.

Server

  • Endpoint: https://app.radi8.com/api/mcp
  • Transport: Streamable HTTP, stateless. There is no session id; every request stands on its own.
  • The server sends instructions for its tools in the initialize result. Clients that pass them to the model get better results.

Sign-in (OAuth 2.1)

  • A request without a token gets 401 with WWW-Authenticate: Bearer resource_metadata="https://app.radi8.com/.well-known/oauth-protected-resource/api/mcp", scope="mcp".
  • Protected resource metadata: https://app.radi8.com/.well-known/oauth-protected-resource/api/mcp. Authorization server metadata: https://app.radi8.com/.well-known/oauth-authorization-server.
  • Dynamic client registration (RFC 7591) at https://app.radi8.com/api/mcp/register. Clients are public: token_endpoint_auth_method is none.
  • Authorization code grant with PKCE; S256 is required. Authorization codes expire after 10 minutes, and the redirect_uri must match the registered one exactly.
  • One scope: mcp. Tokens are bound to the resource https://app.radi8.com/api/mcp (RFC 8707); the resource parameter is optional and defaults to it.
  • A token works until the person revokes it in Radi8.

API keys

  • For a client that cannot run the browser sign-in (a script, a server-side agent, a homegrown client), create an API key in the Radi8 app: Settings, then Connect to Chat, then Your API keys. The key is shown once; it starts with radi8_.
  • Send it as a bearer token: Authorization: Bearer radi8_.... It acts as the person who created it, with the same access as the OAuth flow.
  • A person can hold up to 10 keys. Every key and every OAuth connection is listed under Your API keys, where it can be revoked. Revoking one does not affect the others.
  • Treat a key like a password: it can schedule posts to public accounts.

Workspaces

  • One token is one Radi8 user. A user has a personal workspace and may administer company workspaces.
  • A new token has no active workspace. Until one is chosen, content tools return an error result that says so. Call list_accounts, ask the person which workspace to use, then call set_active_account. The choice is saved on the token.
  • Only company workspaces the person administers can be selected.

Tools

Hints are the annotations each tool declares: read (readOnlyHint), destructive (destructiveHint), open-world (openWorldHint: reaches outside Radi8, to a social channel or a public URL).

Workspaces

For example: “List my Radi8 workspaces and switch to my company.”

list_accounts read
get_active_account read
set_active_account write

Brand context and strategy

For example: “Check what Radi8 knows about my brand and ask me about anything missing.”

get_brand_context read
update_brand_context writedestructive
get_strategy read
update_strategy writedestructive

Monthly plan

For example: “What is in this month's plan that I have not posted about yet?”

get_plan read
get_plan_items read
update_plan writedestructive
create_plan_item write
edit_plan_item writedestructive
delete_plan_item writedestructive
create_campaign write

Drafts

For example: “Draft a LinkedIn post in my voice about what we shipped this week.”

list_drafts read
create_draft write
edit_draft writedestructive
delete_draft writedestructive

Scheduling

For example: “Write a version of that for X and schedule it for next Tuesday at 9am.”

get_connected_channels read
list_scheduled_posts read
schedule_post writedestructiveopen-world
schedule_draft writedestructiveopen-world
edit_scheduled_post writedestructive
unschedule_post writedestructive
delete_scheduled_post writedestructive

Published posts and comments

For example: “Add the article link as the first comment on my latest LinkedIn post.”

list_published_posts read
get_post_comments readopen-world
create_comment writedestructiveopen-world
reply_to_comment writedestructiveopen-world

Media

For example: “Attach the product photo from my media library.”

get_media_library read
upload_image writeopen-world

Brand kit

For example: “Build my brand kit from this screenshot of our homepage.”

get_brand_kit read
update_brand_kit writedestructive
set_brand_kit_asset writedestructive
extract_brand_kit_from_image writedestructiveopen-world

Saved inspiration

For example: “Show me what I saved in Discover that I could write about.”

get_discover_items read

Limits

  • Posts publish at their scheduled time; there is no publish-now. Scheduled times must be in the future.
  • Comments and replies on LinkedIn post right away.
  • Nothing already published can be edited or deleted.
  • No image or video generation. Images come from a public URL (upload_image) or the person's media library; raw image bytes are not accepted.

Errors

  • Tool failures are tool results with isError: true and a message written for the model, not JSON-RPC errors. Pass the message to the model: it says what to do next.
  • A new account comes with credits to try Radi8. Once they are used up, tools that need a subscription return an isError result that links to billing in the app.
  • A rejected token gets 401 with error="invalid_token" or error="insufficient_scope" in WWW-Authenticate; start the sign-in again.
  • Requests are rate-limited. On 429, wait and retry.

Your data

  • When you use the connector, the workspace data your assistant asks for is sent to the provider of that assistant so it can answer you, the same as if you pasted it into the chat yourself. Radi8 sends only what the request needs; what the provider does with it is governed by their terms.

The privacy policy covers how Radi8 handles your information generally.

Help

Email support@radi8.com or see support. Setup for people: the install guide.