Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.auditynow.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Requirements

  • ChatGPT account on Plus, Team, or Enterprise (Custom GPTs require a paid plan)
  • An Audity Personal Access Token (aky_...), see Authentication

​
Setup

The agent surface is exposed as an OpenAPI spec. ChatGPT imports the spec and generates one Action per endpoint.
1

Create a Custom GPT

ChatGPT → sidebar → Explore GPTs → Create.
2

Configure the basics

Give it a name (Audity), a one-sentence description, and a short instruction telling it what Audity is and when to use it.Sample instruction:
You have access to Audity, a B2B SaaS platform for AI consultants. Use it whenever the user asks about audits, audit projects, AI readiness scores, leads from ReadyLinks, or their Nucleus memory layer. Ask the user before triggering audit analysis (it costs 1,000 credits) or converting batches of leads.
3

Add the Action

Configure → Actions → Create new action.
4

Import the OpenAPI spec

In the Schema field, click Import from URL and paste:
https://docs.auditynow.com/api-reference/openapi.json
ChatGPT will fetch and parse it. You should see ~26 operations listed.
If the import fails (Mintlify occasionally serves the spec with strict CORS), download the JSON locally and paste it directly into the schema field.
5

Set authentication

Authentication type: API KeyAPI key: paste your aky_... token (just the token, no Bearer prefix, ChatGPT prepends it automatically based on the auth type).Auth Type: BearerCustom Header Name: leave blank (defaults to Authorization).
6

Set the privacy policy URL

Required by ChatGPT before you can save: https://auditynow.com/privacy
7

Test in the preview

In the right-hand preview pane, ask:
List my Audity projects.
ChatGPT will pick the listProjects action, call the API, and return your real projects. If you see “Allow” prompts, approve them, Custom GPTs ask for confirmation on every Action call until you check “Always allow.”

​
Authentication, more detail

Custom GPT Actions handle the Authorization: Bearer <token> header for you when:
  1. Auth type is API Key
  2. Auth Type subselect is Bearer
  3. You paste the raw token (no Bearer prefix yourself)
If you paste Bearer aky_... instead of just aky_..., ChatGPT will send Authorization: Bearer Bearer aky_..., every request returns 401 PAT_MALFORMED. This is the most common ChatGPT setup failure.

​
Common prompts

Same prompt menu as Claude, both are reading the same spec. 
What's the status of my latest Audity audit?

Pull the deliverables for project {id}. Format the executive summary as a
client-ready email.

Search my Audity Nucleus memories for {topic}.

Show me my unread Audity insights, grouped by type.

​
Limits ChatGPT imposes

A few things to know before you scale this:
  • One Custom GPT, one OpenAPI spec. No multi-spec composition.
  • GPTs ask for permission on every Action call unless you explicitly mark a domain as “always allow.” For workflows that chain 5+ calls, that’s a lot of clicks. Use Claude or Cursor for chained workflows.
  • No streaming responses from Actions. Audity’s responses are JSON, so this doesn’t matter for our endpoints.
  • Schema URL is fetched at import time, not at runtime. If you update Audity’s spec, you’ll need to re-import. Until we ship a webhook for spec changes, treat the import as a frozen snapshot.

​
Troubleshooting

Mintlify occasionally serves the spec with strict CORS. Download the spec locally:
curl https://docs.auditynow.com/api-reference/openapi.json -o audity-spec.json
Then paste the JSON contents into the schema field instead of the URL.
You probably included Bearer in the API key field. Edit the Action’s auth, paste just aky_<32-chars> with no prefix.
Your token lacks the write scope. Generate a new token with both read and write, then update the Custom GPT’s API key.
Custom GPTs can be conservative about external Actions. Make sure your prompt mentions “Audity” by name. Tighten the GPT’s system instruction with examples of when to call which Action.
See Authentication → Rate limits. The 30-writes-per-minute limit is the most common ceiling for ChatGPT batch workflows.
POST /api/projects/{id}/audit-analysis can run 60–300 seconds. ChatGPT Actions have a hard timeout that’s shorter than that, sometimes the call succeeds server-side but ChatGPT reports a failure. Verify with GET /api/projects/{id}/audit-analysis after a minute or two before re-triggering.

​
What’s next