Customgpt

@customgpt what are my current usage limits and how many agents do I have left this month?

Based on your CustomGPT account, you're currently using 3 of your 5 available agents. You've consumed 12,400 of your 50,000 monthly word quota (24.8% used) and have processed 340 of 1,000 query credits (34% used). Your billing cycle resets in 18 days. All limits are within normal range for your plan tier.

This example shows the MCP's ability to surface account-level quotas in real time. Requires a valid API key with read access to user profile endpoints. Useful for capacity planning before deploying new agents or scaling usage across your team.

@customgpt retrieve the full conversation history from project ID proj_8x2k9 so I can review how users are interacting with our support bot

Fetched 47 messages from conversation thread in project proj_8x2k9. The exchange shows a customer asking about return policies (3 messages), followed by the agent providing policy details and processing a return request (2 messages), then troubleshooting a shipping label issue (5 messages). The conversation spans 22 minutes with an average response time of 8 seconds. Most recent message was 14 hours ago.

Demonstrates read-heavy retrieval of chat logs for a deployed CustomGPT agent. You must have the correct project ID and appropriate license access. Conversation data can be large; consider filtering by date range if the thread is long-running to avoid token overhead.

@customgpt delete the staging agent proj_test_4a7 and its associated license lic_9f3x since we've migrated to production

Successfully deleted CustomGPT project proj_test_4a7 and removed license lic_9f3x. The agent is no longer accessible via API or web interface, and all associated conversation history has been archived per your account's retention policy. Quota for this agent (1 of 5 slots) is now freed and available for new projects.

Shows a destructive write operation with permanent side effects. Always verify project and license IDs before deletion—this action cannot be undone. The MCP confirms both the project removal and license cleanup, which is critical for maintaining accurate billing and quota tracking.

When you need to trace what your support bot actually said

A 6-person support team runs a CustomGPT agent that answers tier-1 questions from their knowledge base. Every morning, the team lead uses List Conversation Messages to pull the previous day's chat transcripts and spot patterns—repeated questions that need better docs, or edge cases where the bot gave a wrong answer. The Get Project Settings tool lets them verify which sources the bot is actually using after they update the knowledge base. This MCP is the right call if you're already paying for CustomGPT and need programmatic access to conversation logs for QA or training data. It's overkill if you're just building a one-off chatbot prototype—use the CustomGPT dashboard instead. If your team needs to audit bot conversations at scale and tie them back to specific agent configs, this MCP closes the loop.

When each customer gets their own branded AI agent

A 12-person B2B SaaS company sells white-labeled help centers, and each customer gets a CustomGPT agent trained on their own docs. The ops team uses this MCP to automate agent lifecycle: Get Current User Profile confirms API auth, then they spin up new projects and licenses via the API when a customer signs up. Delete CustomGPT Project and Delete CustomGPT Project License handle churn. Get Usage Limits runs nightly to flag accounts approaching quota so they can upsell or throttle. This MCP is essential if you're reselling CustomGPT as part of your product and need to manage dozens or hundreds of agents programmatically. It's not worth the setup if you're running fewer than 10 agents—just use the web UI. If your business model is 'one agent per customer' and you need to automate onboarding and teardown, this MCP is the infrastructure play.

When you need to prove which docs your bot used last quarter

A 20-person fintech startup uses CustomGPT to answer internal compliance questions. Every time they update a policy doc, they need to audit which version of the knowledge base was live when specific answers were given. The compliance lead uses Get Project Settings to snapshot the agent's source configuration at the end of each month, and List Conversation Messages to pull Q&A logs tied to that snapshot. This creates an audit trail for regulators. This MCP is the right tool if you're in a regulated industry and need to reconstruct exactly what your bot knew at a point in time. It's unnecessary if your knowledge base changes rarely or you don't face external audits. If you need to prove 'the bot only had access to approved docs on this date,' this MCP gives you the receipts.