The Typetone MCP server lets an AI assistant work with your Typetone workspace. From Claude, ChatGPT or a coding tool you can check a draft against your rules, start a compliance audit, and review the violations an audit found, without opening the Typetone app.
This page is the technical reference. For a general introduction, see API, MCP and using Typetone in your AI tools.
Server details
| Server URL | https://api.app.typetone.ai/mcp/ |
| Transport | Streamable HTTP |
| Authentication | OAuth 2.1 with PKCE, or an API key |
| Access | Workspaces on the Enterprise plan |
| Scope | One workspace: the one you sign in to, or the one your API key belongs to |
Connect Claude, ChatGPT and other AI apps
AI apps that support custom connectors sign you in with OAuth. You don’t need an API key.
- Add a custom connector in your AI app. In Claude, go to Settings → Connectors.
- Paste the server URL:
https://api.app.typetone.ai/mcp/ - When asked, sign in with your Typetone account.
Everything the assistant does then runs as you. Review decisions and comments are saved under your name, just like when you work in the app.
Connect Claude Code, Cursor and Codex
Coding tools can send an API key as a header. Create a key in the Typetone app under Settings → MCP & API. That page also generates a ready-to-paste install command for Claude Code, Cursor, Codex and Gemini CLI.
claude mcp add --transport http typetone https://api.app.typetone.ai/mcp/ \
--header "Authorization: Bearer YOUR_API_KEY"
Send the key as Authorization: Bearer <key> or as X-API-Key: <key>. Treat it like a password: keep it out of shared code and revoke it in the app when you no longer need it.
Actions taken with an API key are recorded for the workspace, not for a specific person. Reviewers who label violations should connect with OAuth so decisions carry their own name.
Authentication details
For client developers and connector directories:
- OAuth 2.1 authorization code flow with PKCE (S256). Public clients only (
token_endpoint_auth_method: none). - Dynamic client registration is supported.
- Protected resource metadata:
https://api.app.typetone.ai/.well-known/oauth-protected-resource/mcp - Authorization server metadata:
https://api.app.typetone.ai/.well-known/oauth-authorization-server - Scopes:
mcp:connectandoffline_access. - Access tokens are valid for one hour. Refresh tokens are valid for 30 days and rotate on every use.
- An unauthenticated request returns
401with aWWW-Authenticateheader that points to the metadata above. - Access is checked on every request: a user who leaves the workspace, or a workspace that leaves the Enterprise plan, loses access right away.
Tools
| Tool | What it does | Changes data | Uses credits |
|---|---|---|---|
list_brand_voices | Lists the brand voices in the workspace | No | No |
get_brand_voice | Returns the guidelines of one brand voice | No | No |
list_content_rules | Lists the compliance and brand rules content is checked against | No | No |
check_content | Checks a text against the workspace’s rules and returns violations with a recommended fix | No | Yes |
get_credit_balance | Returns the remaining credit balance | No | No |
list_rulesets | Lists the rulesets (rule groups) available for audits | No | No |
list_assets | Lists or searches assets in the Asset Vault | No | No |
audit_assets | Starts a compliance audit of up to 10 assets against one ruleset | Yes | Yes |
get_audit_results | Returns the progress and violations of an audit | No | No |
list_audits | Lists the workspace’s audits, newest first | No | No |
list_violations | Pages through the violations of one audit, filtered by review state, severity or search term | No | No |
get_violation | Returns one violation in full: flagged text, reason, rule, recommended fix, suggested text and comments | No | No |
review_violation | Records a review decision on a violation | Yes | No |
comment_on_violation | Adds a comment to a violation’s thread. Mentions in the comment don’t assign the violation; assign in the app. | Yes | No |
Review an audit from your AI assistant
Compliance reviewers can work through an audit in a conversation instead of in the app. A typical session:
- “Show me my latest audits.” The assistant calls
list_audits. - “Let’s review the undecided violations of the one from Monday.” The assistant calls
list_violations, which shows undecided violations first. - For each violation the assistant shows the flagged text, why it was flagged and the recommended fix (
get_violation). - You decide, and the assistant records it with
review_violation.
review_violation takes one of these actions:
| Action | Meaning |
|---|---|
dismiss | Not a real violation. A reason is required, and it’s saved with the decision. Other actions don’t take a reason: add a comment instead. |
elevated_risk | A real violation with elevated risk. |
urgent_violation | A real violation that needs action now. |
mark_fixed | The fix has been checked and the violation is resolved. |
reopen | Back to undecided. Clears an earlier decision. |
Set apply_to_group to true to apply the same decision to every violation with the same group_id: the same finding on several assets.
Decisions follow the same rules as in the app. They show up right away in the audit’s review mode, with the reviewer’s name in the activity trail. A violation that is assigned to a colleague can only be dismissed or marked fixed by that colleague, the person who assigned it, or an admin.
Credits and limits
check_contentandaudit_assetsuse workspace credits. Reading data and recording review decisions is free.audit_assetsasks for confirmation when an audit’s estimated cost is large. The assistant then repeats the call with the returnedestimate_id.- Credit-using tools are limited to 20 calls per minute and 3 at a time per workspace.
check_contentaccepts up to 60,000 characters per call.list_violationsreturns up to 50 violations per page.
REST API
The same review actions are available over the REST API, for your own scripts and systems. Base URL: https://api.app.typetone.ai/public/v1. Send your API key in the X-API-Key header.
| Method | Path | What it does |
|---|---|---|
GET | /audits | List audits, newest first |
GET | /audits/{audit_id}/violations | List violations, with review_state, severity, search, page and page_size |
GET | /violations/{violation_id} | Get one violation with its comments |
POST | /violations/{violation_id}/review | Record a decision: {"action": "dismiss", "reason": "..."} |
POST | /violations/{violation_id}/comments | Add a comment: {"comment": "..."} |
The full, interactive reference is at api.app.typetone.ai/public/docs.
Example prompts
- “Check this LinkedIn post against our brand and compliance rules, and fix what’s wrong.”
- “Does this landing page copy contain green claims we can’t back up?”
- “Audit the three newest pages in our Vault against our marketing compliance ruleset.”
- “Walk me through the undecided violations of our latest audit, most severe first.”
- “Mark this one as an urgent violation and add a note that legal needs to see it.”