AI
Native, not bolted on.
AI in Requester isn’t a chat box in the corner. It drafts the request, builds the form, suggests the approvers, summarizes for the reviewer, chases the stalled ones, and writes the record at the end — inside the same system that holds the data.
Free plan, no card required. Submitters are always free.
Rex
What’s waiting on me, and what’s about to breach?
reading pending_approvals, workspace_sla
Four reviews are waiting on you. Two breach the workspace SLA tomorrow:
- FIN-284MacBook Pro — $2,48018h left
- IT-119Figma seat — Design22h left
Most tools added a chat box. This one reads the workspace.
An assistant that can't see your pending approvals, your templates, or who's on leave can only ever rewrite your sentences. Rex has scoped access to the actual objects.
- ~20
- Scoped tools Rex uses to read your workspace — requests, approvals, templates, workflows, members, usage
- 4
- AI providers supported: Azure OpenAI, OpenAI, Anthropic, and Google
- 5
- Agent archetypes: steward, reviewer copilot, intake, workflow step, coordinator
- 0
- Hard stops — paid workspaces degrade to a cheaper model rather than blocking
The assistant that already knows what you’re asking about
Ask
“What’s waiting on me, and what’s about to breach?”
Streaming chat with history, titles, and summaries — reading pending approvals, recent requests, activity, comments, templates, workflows, teams, and workspace usage through scoped tools.
- Context-aware prompt suggestions on the dashboard, so it's useful before you know what to type
- Drafts a custom request end to end from a description
- Sees your active draft, so “finish this for me” means something
What Rex can reach
- Approvals
- Pending decisions, urgency, who else is assigned
- Requests
- Details, answers, activity, comments
- Structure
- Templates and fields, workflows, teams, groups
- People
- Members, invitations, your own profile
- Usage
- Workspace stats — requests, storage, seats, AI pools
Generate
Forms and workflows from a description
Describe what you need to collect and how it should be approved. Requester generates both — streaming as it builds, validating the workflow graph before saving, and leaving everything editable.
- Form generation with the right field types and validation inferred
- Workflow generation with branches, approvers, and reminders
- Post-processing validates the output instead of trusting it
Prompt
“Vendor onboarding — legal reviews the contract, finance sets payment terms.”
Form
9 fields inferred, including attachment and currency types
Workflow
2 approval steps, 1 conditional branch, 1 SLA reminder
Validated
Graph checked and normalised before it's saved
Four places it earns its keep every day
Not demos — the moments where a human would otherwise spend fifteen minutes reconstructing context that the system already has.
Routing suggestions
At submit time, who should review this and why — with the reasoning shown, so you can disagree with it rather than just accept it.
Request summaries
A plain-English précis of every submission, generated automatically and carried into the reviewer's email as well as the app.
Decision narratives
When a request closes, the AI writes a readable account of how it got decided from the actual event history — not a template.
Genesis onboarding
Describe your company and an agent proposes a starting workspace — teams, templates, workflows, policies — that you apply in one step.
Something to chase the requests you’ve stopped looking at
The steward archetype works the stalled ones. Others sit alongside reviewers, handle intake, run as a workflow step, or coordinate across several. Each is activated per workspace and gated by an explicit mandate and autonomy setting.
Steward
Chases reviews that have stalled past their window.
Reviewer copilot
Sits with an approver and answers questions about the request.
Intake
Turns an unstructured ask into a well-formed request.
Workflow step
Runs as a node inside a workflow execution.
Coordinator
Works across several agents on one request.
Every agent run is recorded and emits audit events. An agent is a participant in your workspace, not a black box acting on it.
Your key, your models, your ceiling
AI pricing that punishes usage teaches people not to use it. Requester pools credits across the workspace, accounts for who triggered what, and degrades instead of stopping.
Bring your own key
Attach your own provider key per workspace and your AI usage is unmetered by us. Your prompts, your contract, your bill.
Four providers
Azure OpenAI, OpenAI, Anthropic, and Google, behind one model catalog with per-tier routing. Swap providers without touching the product.
Pooled credits
Credits are sized per seat and shared across the whole workspace, so one heavy user doesn't starve everyone else out.
Degrade, never stop
Past the pool, paid workspaces keep working on a faster, cheaper model at no extra charge. An approval never blocks on a meter.
Three payers, shown separately
User-triggered
AI a member asked for directly — drafting, chat, generation.
Workspace automation
AI your own workflows and agents triggered without a human in the loop.
Vendor-paid onboarding
Genesis and setup assistance, on us, not billed to your pool.
It drafts, summarizes, suggests, and chases. People decide — and the audit log records which was which.
On AI, cost, and control
- Does the AI approve things on my behalf?
- No. It drafts, summarizes, suggests, and chases — humans decide. Agents operate under an explicit mandate and autonomy setting per workspace, every run is recorded, and their actions land in the audit log like anyone else's.
- Can I use my own OpenAI or Anthropic key?
- Yes. BYOK is per workspace and covers Azure OpenAI, OpenAI, Anthropic, and Google. With your own key attached, your AI usage is unmetered on our side entirely.
- What happens when we run out of AI credits?
- On a paid plan, nothing breaks — you keep working on a faster, cheaper model at no extra charge. Credits are pooled across the workspace and sized per seat, and the billing page shows pool usage, your share, and automation broken out by source.
- Who pays for AI an automation triggers?
- It's accounted separately. Three payers are tracked distinctly: AI a user triggered, AI the workspace's own automation triggered, and vendor-paid onboarding. You can see which is which rather than getting one opaque number.
- What can the assistant actually see?
- Rex reads your workspace through roughly twenty scoped tools — pending approvals, recent requests, request details and activity, comments, templates and their fields, workflows, teams, groups, members, usage stats, invitations, and your active draft. It's scoped to the workspace you're in and the permissions you already have.
The rest of the product
Intake & forms
27 field types, conditional logic, and six ways a request can arrive.
Approvals
Any, all, or N-of-M — decided in-app, from email, or by outside guests.
Workflow automation
A visual canvas for routing, branching, delays, SLAs, and webhooks.
Tracking & audit
Dashboards, analytics, a full audit log, and shareable status pages.
Integrations
Slack, Teams, Discord, Google Chat, webhooks, Zapier, and the API.
Describe your process. Have it running before the meeting ends.
The free plan includes a pooled AI allowance — enough to generate your first templates and workflows and see whether the thing actually understands your business.