Skip to main content

The Creopus Agent

Video transcript

00:00 — The Creopus agent

The Creopus agent. Walkthrough of the agent panel.

00:07 — Your sessions

Each conversation with the agent is a session that persists. Click any session to resume it; the agent remembers the full message history.

00:19 — Start fresh

Click plus New for a clean session. Useful when you switch context — a different system, node, or problem — so the agent doesn't carry over memory that no longer applies.

00:36 — Tune the output style

Pick concise for tight answers, normal for the default balance, or verbose when you want the agent to show its reasoning. Applied on the next turn.

00:49 — Daily token budget

Live cost and tool-call count for the session, alongside your daily token budget. The Free plan has no agent access. Paid plans get 200 thousand tokens a day on Pro, 1 million on Team, and 10 million on Enterprise — resetting at midnight UTC.

01:12 — Talk to the agent

Type a request in plain English. Audit my BMS requirements against ISO 26262. Draft a DFMEA for the charger module. Type / for commands, or @ to pull a hierarchy node or artifact into context. And when the agent changes a document, that change always arrives as a follow-on you review and merge — never a silent edit.

01:47 — docs.creopus.ai

Full documentation, including every tool, is at docs.creopus.ai

What it does

The Creopus Agent is an autonomous reasoning loop that drives the platform's own tools to complete multi-step engineering work from a single plain-English request. Instead of you opening ReqGen, then DFMEA, then the BOM by hand, you ask once — "audit my BMS requirements against ISO 26262 ASIL-D and raise a DFMEA for the gaps" — and the agent chains the tools together, grounded in your System Hierarchy.

It reaches the same tools you use by hand and reads the same hierarchy, so its output lands in exactly the same place as your own — as artifacts on nodes, subject to the same review and change control.

Why it matters

The value of an all-in-one platform is that every tool sees the same source of truth. The agent is what turns that into leverage: a request that would be an afternoon of tool-hopping and copy-paste becomes one instruction, and because the agent works inside the platform rather than beside it, nothing it produces has to be re-imported or reconciled. It is the difference between an assistant that drafts text you paste back in, and one that operates the system.

How it works

  1. Ask in plain English. Type a request — "draft a DFMEA for the charger module", "find inconsistencies between my requirements and the BOM". Use / for slash commands and @ to pull a specific hierarchy node or artifact into context.
  2. The agent plans and acts. It selects and calls the platform's tools in sequence, reading requirements, DFMEA, block diagrams and BOM from the relevant node to ground each step.
  3. You review, then merge. When the agent changes a document, the change never lands silently — it always arrives as a follow-on you inspect in the diff panel and merge, exactly like a change proposed by a teammate.

Sessions

Each conversation is a session that persists with its full message history. Resume any session to pick up where you left off, or start a clean one when you switch context — a different system, node or problem — so the agent doesn't carry over memory that no longer applies.

Output style

Tune how much the agent shows you, applied from the next turn:

  • concise — tight answers, bullets over paragraphs
  • normal — the default balance
  • verbose — the agent shows its reasoning

On every tool page

The same agent also lives as a docked panel on every tool page, pinned to the document you have open. The Ask agent and Draft with the agent buttons pre-brief it on that document, so you can delegate a single slice — "add three thermal failure modes" — without re-explaining what you are working on. As on the dedicated page, any change it makes returns as a follow-on for you to review and merge.

The follow-on contract

The agent is never permitted a silent edit. Everything it changes on a document becomes a follow-on that you review against the diff and merge — the same change-controlled loop that governs your own edits and a reviewer's. That is what makes an autonomous agent safe to point at controlled engineering documents: its work is proposed, not applied, and you remain the one who commits it.

Daily token budget

Agent usage is metered by a daily token budget that resets at UTC midnight. The panel shows live cost and tool-call count for the current session against your remaining budget.

PlanDaily agent budget
FreeNo agent access
Professional200K tokens/day
Team1M tokens/day
Enterprise10M tokens/day

Beyond the browser

The same tools the agent drives are also exposed to external AI clients through the MCP server, so you can reach Creopus from Claude, Cursor and other assistants. The in-platform agent and the MCP surface share one tool catalog, so they never drift apart.