PROMIVO Co-Pilot
The command deck for work no single agent covers
- 8hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 8output fields
- A structured result, not a paragraph you have to re-read
- ≤8steps per run
- Bounded work, enforced by the runtime rather than requested in the prompt
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
Every other agent in this catalogue is a specialist. That is what makes them reliable, and it is also what leaves a gap: most of a working day is not one narrow task. It is reading a document that arrived this morning, working out what actually matters in it, deciding what to do next, and putting that into a form somebody else can act on.
The Co-Pilot is the agent for that work. Give it a contract, a report, a set of notes, a result another agent produced, and ask it a real question about them. It answers from the material rather than from general knowledge, says plainly when the material does not contain the answer, and separates what the source says from what it is inferring.
It also knows the rest of the catalogue. Describe a recurring job and it can tell you whether a specialist already exists for it, what that agent needs as input, and what it returns — which is knowledge no general assistant has.
What it will not do is run that agent for you. It holds no tool that executes another agent, sends anything, publishes anything or contacts anyone; its risk ceiling is read-only and enforced by the runtime rather than by its own prompt. Recommending and acting are different decisions, and the second one stays with you.
It runs on your own AI provider key. Model usage is billed to you by your provider at their prices, with nothing added.
Benefits
The work between the specialists
The reading, deciding and drafting that no single-purpose agent covers.
A route into the catalogue
Describe the problem in your own words and find out whether something here already solves it.
Nothing happens without you
Read-only by construction, so the worst outcome is advice you can reject.
Billed by your provider, not by us
It runs on your own model key at your provider's prices.
What it does
Answers from your material, not from memory
Attach documents and ask about them. What the source does not say, it says it does not say, rather than filling the gap from general knowledge.
Knows what else is in the catalogue
Describe a recurring job and it can name the specialist agent that already does it, what it needs, and what it returns.
Separates fact from inference
What the document states, what follows from it, and what is an assumption are marked as three different things.
Structured output for business use
Summary, findings, what is missing, and a recommended next step — rather than one long paragraph to re-read.
Reads what the platform can read
PDF, DOCX, CSV, TXT and Markdown, through the same extractor the rest of the platform uses. A truncated document is named as truncated.
Recommends, never executes
It holds no tool that runs another agent, sends, publishes or contacts anyone. The risk ceiling enforces that independently of its prompt.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3ToolsNo external tools
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataAn illustration of the shape, not a real agreement. Your findings come from the document you supply.
Input
{
"task": "Read this supplier agreement and tell me what we are committed to that has a date attached.",
"documents": [
{
"name": "supplier-agreement.pdf",
"text": "…extracted contract text…"
}
]
}Output
{
"missing": [
"The agreement names no owner for the quarterly report.",
"No notice address is given for the insurance evidence, so it is not clear who it goes to."
],
"summary": "Four dated commitments run to you under this agreement, and one of them is a reporting obligation with no owner named anywhere in the document.",
"escalate": false,
"findings": [
{
"kind": "stated",
"basis": "supplier-agreement.pdf, clause 7.2",
"statement": "A quarterly service report is due within 15 business days of each quarter end."
},
{
"kind": "stated",
"basis": "supplier-agreement.pdf, clause 11.4",
"statement": "Insurance must be maintained at £5m and evidenced annually on the anniversary of signature."
},
{
"kind": "inferred",
"basis": "Follows from clauses 7.2 and 11.4 read together.",
"statement": "The first report therefore falls due before the first insurance evidence date."
}
],
"nextStep": "Assign an owner for the quarterly report before the first quarter end, and confirm the notice address with the supplier.",
"disclaimer": "Analysis of the document you supplied, not legal advice, and not a record of anything performed. Nothing has been sent, filed or run. Where the agreement is silent this says so rather than inferring a term.",
"suggestedAgent": {
"why": "This is a recurring job across a portfolio of contracts rather than a one-off read, and that agent extracts dated obligations with owners and deadlines across all of them.",
"name": "Contract Obligation Tracker",
"slug": "contract-obligation-tracker",
"neededInput": "The signed agreements, and the date you want the schedule calculated from."
}
}Requirements
Integrations
No integrations required.
Before you start
- A connected AI provider key — the agent runs on your own account.
- The material you want analysed, as text or as a file the platform can extract.
- Requires the All-Access plan or higher
Use cases
Read this and tell me what matters
A contract, a supplier report, a policy update. What it says, what follows from it, and what it leaves unanswered.
Which agent should I use for this?
Describe a recurring job and find out whether a specialist already exists, and what it would need from you.
Make this usable by someone else
Turn working material into a summary, a structured brief, or a set of next steps.
Help me read an agent's result
Take the output of a specialist run and work out what it means and what to do about it.
Prepare an input
Work out what a specialist agent needs and get the material into that shape first.
Pricing
Free
Free to deploy. Runs are metered against your plan allowance.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Can it run other agents for me?
No. It can tell you which agent fits and what it needs, but it holds no tool that executes one. Running an agent stays a decision you make, from that agent's own page. This is a hard limit in the runtime, not a preference in its instructions.
Does it use my own AI key?
Yes. Like every agent here it runs on your connected model key, and your provider bills you directly at their prices. PROMIVO never runs an agent on its own key.
What files can it read?
Whatever the platform's extractor supports — PDF, DOCX, CSV, TXT and Markdown. Files are parsed by the existing extraction endpoint, and anything it cannot read is refused there rather than reaching the model as noise.
Will it make something up if my document does not cover it?
It is instructed to say what the material does not contain, and to mark inference separately from what a source states. No safeguard makes a language model incapable of error, so treat its output as analysis to check rather than fact to file.
Can a document tell it to ignore its rules?
Document text is treated as evidence, not as instruction. A file containing commands is reported as a file containing commands.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent