Enterprise
Comprehensive AI solutions and support for executive outcomes at scale.
Enterprise AI PlatformLearn
Turn your prompt into a multi-task plan.
Instantly search your entire company knowledge.
Analyze documents, PDFs and spreadsheets.
Analyze documents, PDFs and spreadsheets.
Interact with any platform using MCPs.
Search the Internet.
Learn more about Enterprise Agent in our docs.
An Enterprise Agent is a CustomGPT.ai agent running in Plan & Act mode. Unlike a standard agent that only answers questions, it breaks down a request into steps and executes them end to end. It reads your files, runs calculations, pulls live data from your existing systems, and delivers a finished output in a single response. No code required to build or deploy. Learn more
Whole tasks, not just answers, across a range of teams. Accounting teams consolidate quarterly P&L statements into one report with year-over-year trends. Legal teams compare contract versions and get a summary of every material change. Support leads turn a week of tickets into a categorized summary. Procurement teams compare vendor bids side by side. Universities give students rubric-based feedback on their essays. It reads the source material in full, does the work, and reviews its own output before handing it back.
Yes, and that’s where most of the value shows up. It can pull a record from an external system, check it against an uploaded file or your knowledge base, and produce the output, all in one pass instead of three separate tools and manual handoffs. For example, it can take an RFP a prospect just sent you, match each requirement against your library of past proposals, and draft a tailored response in one sequence. Learn more
A finished deliverable. When the output is too large or structured for a chat window, it creates a downloadable file or builds an interactive dashboard you can filter and explore, right in the conversation.
Yes. With Make External Connections enabled, the agent can interact with any platform using MCPs, plus 100+ pre-configured integrations including HubSpot, Pipedrive, Zendesk, Freshdesk, Intercom, Jira, Asana, ClickUp, Stripe, and GitHub. Anything not on that list connects through a Custom API or MCP connector. Learn more
Both. An Enterprise Agent is something you deploy, not just use: on your website as an embedded widget, behind a share link, behind a login for a specific group, or inside your product under your own brand.
Your current agent answers from what it knows and can read an attached file or reach a few outside systems. Plan & Act adds a planning step: the agent breaks a request into steps and works through them in sequence, adapting as it goes. Generating a downloadable file, building a live dashboard, and chaining several tools together in one response are things a regular agent can’t do.
Those are personal assistants designed for one person in one chat window, inside someone else’s product. An Enterprise Agent is infrastructure you own. It answers from your knowledge base with citations back to your source documents, follows the persona and guardrails you set, and can be deployed to your customers under your own brand. You are not locked to a single AI provider either.
It plans the steps and works through them on its own, you don’t prompt it turn by turn. It only stops when it genuinely doesn’t have what it needs, a missing file, a needed clarification, and it tells you rather than guessing or delivering an incorrect result with confidence.
No. Turn on Plan & Act from Personalize > Intelligence, or from the three-dot menu next to your agent under Actions. Connecting a platform means selecting it from a list and entering an API key. Learn more
Yes. Your persona and guardrails stay active for the entire task. The planning step operates inside the rules you set as the builder, it can’t override them, and neither can instructions from your end users.
Yes. API keys are encrypted at rest in a vault, and the agent never touches them directly. It works with a short-lived temporary key instead, swapped in through a secure background process, so exposing it mid-conversation couldn’t expose your real credentials. Learn more
Yes. On Enterprise you can set per-user query-credit limits, daily or monthly, for an individual, a role, or guests on a public deployment, under My Profile > Teams > Usage Limits. When someone hits their limit, the agent stops responding until it resets, with a notice you can word yourself. Learn more
Standard, Premium, and Enterprise all include it. Running it costs credits based on what the agent actually does, planning the task, checking it’s complete, and running any actions. See the full credit breakdown
It reads pdf, docx, doc, odt, txt, jpg, jpeg, png, webp, xlsx, xls, csv, tsv, json, md, and pptx in full, not a sample, and creates md, txt, pdf, docx, csv, html, and more.