canonical: https://jentic.com/apis/customgpt.ai/customgpt

# CustomGPT.ai API

The CustomGPT.ai API builds and operates retrieval-augmented AI agents grounded in your own content. Across 64 endpoints it creates and configures agents, ingests documents and data sources, runs chat completions in OpenAI format, manages conversations and citations, checks message trust scores, and returns traffic, query, and lead analytics.

## For AI agents

Build and configure retrieval-augmented AI agents, ingest documents and sources, run grounded chat completions, and read conversations, citations, and analytics across 64 endpoints. Authenticates with a bearer token.

## Scope

Does not train foundation models or provide general-purpose generation outside an agent's content. Use for building and running content-grounded CustomGPT.ai agents only.

## Capabilities

- Create, configure, clone, and delete custom AI agents
- Ingest documents and data sources and reindex or sync them
- Run chat completions in OpenAI format grounded in your content
- Manage conversations and export their messages
- Retrieve citations and message trust scores for answers
- Pull traffic, query, conversation, and lead analytics

## Use cases

### Agent-Managed Knowledge Base Chatbots

An AI agent connected through Jentic can stand up and maintain a grounded chatbot without a human in the dashboard. It creates a CustomGPT.ai agent, adds documents and website sources, reindexes them, and then routes user questions to the agent's chat completion endpoint so answers stay tied to the ingested content.

Example prompt: Create an agent, add a set of documentation URLs as sources, reindex them, then send a user question and return the grounded answer

### Grounded Answer Verification

When answers must be trustworthy, the source matters as much as the text. The CustomGPT.ai API returns citations, message claims, and a trust score for each answer and can verify a message's accuracy, so an agent can surface the supporting sources and hold back low-confidence responses.

Example prompt: Send a question to an agent, read the answer's trust score and citations, and only return the answer if the trust score is high enough

### Conversation Analytics and Lead Export

Teams want to know what users ask and who engages. The CustomGPT.ai API reports traffic, query, and conversation analytics and exports captured leads, letting an agent compile a weekly summary of top questions and hand new leads to a CRM.

Example prompt: Pull the last week of query analytics for an agent and export any new leads it captured

## Key endpoints

| Method | Path | Description |
| --- | --- | --- |
| POST | `/api/v1/projects` | Create an agent |
| POST | `/api/v1/projects/{projectId}/sources` | Add a data source to an agent |
| POST | `/api/v1/projects/{projectId}/chat/completions` | Send a message in OpenAI format |
| GET | `/api/v1/projects/{projectId}/conversations` | List conversations for an agent |
| GET | `/api/v1/projects/{projectId}/reports/queries` | Get query analytics for an agent |

## Key resources

- **Agents** — Create, view, update, clone, and delete custom AI agents and their settings
- **Sources and Pages** — Add, list, sync, and delete documents and data sources and reindex them
- **Conversations** — Create and manage conversations and send messages and chat completions
- **Citations and Trust** — Retrieve citations, message claims, and trust scores and verify answers
- **Reports** — Traffic, query, conversation, and lead analytics for an agent

## Why Jentic

- **Setup:** Wiring CustomGPT.ai by hand means managing a bearer token and chaining agent creation, source ingestion, reindexing, and chat calls across 64 endpoints. Through Jentic you install once, import CustomGPT.ai from the API Directory, store the token once, and your agent calls the operations directly.
- **Permission scoping:** Almost every operation addresses a project, page, or conversation by its ID in the URL path, so a rule can bound your agent to a single agent project as well as to specific operations. You might allow sending messages and reading citations while blocking agent and source deletion.
- **Credential handling:** Your CustomGPT.ai bearer token is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context.
- **Discovery method:** Agents search Jentic by intent such as 'send a message to a CustomGPT agent', and Jentic returns the matching operation with its input schema so the agent calls it with the right project ID and message.

## Related APIs

- **Chatbase** — Build and deploy custom AI chatbots trained on your own data
- **Dify** — LLM application and RAG agent platform with a visual builder
- **OpenAI** — General-purpose language models and chat completions
- **Pinecone** — Vector database for storing and querying embeddings

## FAQ

### What authentication does the CustomGPT.ai API use?

The CustomGPT.ai API authenticates every request with a bearer token in the Authorization header, in JWT format per its OpenAPI spec. Through Jentic the token is stored once by your own self-hosted instance and injected when the agent calls, so it never appears in the agent's prompt or logs.

### Is there a CustomGPT.ai MCP server?

You don't need an MCP server to give your agent the CustomGPT.ai API. Jentic connects it directly from the API Directory: import it, store your bearer token once, and your agent calls the agent, source, and chat operations. Operations are discovered on demand, so nothing extra loads into the agent's context.

### Can I limit what my agent is allowed to do with the CustomGPT.ai API?

Yes. Almost every operation addresses a project, page, conversation, or other record by its ID in the URL path, so a rule can bound the agent to a single agent project as well as to specific operations. You might allow sending messages and reading citations while blocking agent and source deletion, with every call logged.

### Does the CustomGPT.ai API support the OpenAI chat format?

Yes. The chat completions operation accepts requests in OpenAI format for a given agent, so an integration built for OpenAI chat completions can call a CustomGPT.ai agent whose answers stay grounded in your ingested content.

### What are the rate limits for the CustomGPT.ai API?

The OpenAPI spec does not state fixed rate limits, though it exposes an account limits and usage endpoint you can query. Check the CustomGPT.ai documentation at https://docs.customgpt.ai for current limits before running high-volume ingestion or chat jobs.

### How do I send a message to an agent through Jentic?

Search Jentic by intent, for example 'send a message to a CustomGPT agent', and Jentic returns the chat operation with its input schema. Your agent supplies the project ID and the message and reads back the grounded answer. To run it on your own infrastructure, install Jentic One from its GitHub repo.