Run your company onwhat it already knows
One place for everything your company knows, with agents, apps and workflows running on top of it.
Ask, and the answer comes back with the file it came from.
What did we decide about pricing last quarter?
Reads from more than thirty services your team already uses.
Everything runs on the same context
Files, decisions and connections live in one place. Every surface reads from it and writes back to it, so the workspace knows more after a piece of work than it did before.
- ChatAsk, and it reads the workspace first
- AgentsStanding work, on a schedule
- WorkflowsSteps you can see and change
- AppsTools built from what is already there
- SkillsHow your company wants things done
- ConnectionsThe tools the work already lives in
Chat is one surface. There are thirteen more.
All of them on the same substrate, so none of them starts from nothing.
- Chat
- Answers grounded in the workspace, not in a guess
- Agents
- Standing jobs on a cron, with an approval gate
- Workflows
- Multi-step runs you can read and edit
- Apps
- Internal tools generated from connected data
- Skills
- The way your company wants a job done, reused
- Vault
- Every file, versioned, previewed before it opens
- Memory
- What was decided, and why, kept between sessions
- Connections
- The tools the work already lives in
- Import
- Bring an existing drive or workspace in whole
- Templates
- A workspace that starts already set up
- Dashboard
- What ran, what it touched, what it produced
- Cost
- Credits by member, by agent, by day
- Trash
- Nothing leaves without a way back
- Workspace
- Members, roles, keys, spend caps, audit
Say what you need. It gets built on what you already have.
Describe the tool. Get the tool.
No per-seat licence, no fresh import, no blank spreadsheet.
Build me a tracker for open renewals by owner, with the value and the close date.
- Reads
- Your CRM, through the connection you already added
- Writes
- A file in the vault, versioned like any other
- Updates
- When the source does, without a re-import
- Shares
- With the workspace, under the same roles
Work that runs without you.
Set a job up in chat. It keeps going after you close the tab.
Every step is visible, and every step is yours.
A run is a diagram, not a black box. Put a person in front of any step that should not happen on its own.
The gated branch waits. Nothing goes out while it does.
It is worth more in month three than in week one
Most AI tools start from nothing every session, so the second month costs exactly what the first one did. Here the context is a thing your company owns and keeps adding to.
- Day one
You point it at a drive and two tools.
The files, and who has been touching themThe words your company uses for its own things - Week two
It stops asking what you meant.
What was decided, and the reason that was givenThe shape your reporting is supposed to takePlus everything it already held.
- Month three
New work starts where the last work stopped.
Skills your team wrote down once and stopped repeatingAgents that already know who has to approve whatPlus everything it already held.
Nothing runs where it should not
Your company's context is the most sensitive thing you will put into any tool. These four are enforced in the system, not in a policy document.
- One workspace cannot see another
- Every read and every write is scoped at the database, not in application code that could forget. There is one role, it is never widened, and there is no path around it.
- No credential ever reaches an agent
- Work runs in a sandbox that is handed an id and nothing else. Your keys stay on our side of that line, so a prompt that talks an agent into leaking one has nothing to leak.
- Every credit is a row you can read
- The usage ledger is append-only. A run that gets declined rolls back its own charge, and the workspace can see what was spent, by whom, on what, on which day.
- Untrusted content is data, never instructions
- A file an agent reads and a page it fetches are treated as hostile input. Neither renders alongside your session, and neither carries your credentials anywhere.
One meter, and you set the ceiling
No seats to count and no plan to outgrow. Work costs credits, the rate is the same at every size, and a workspace stops at the cap you gave it instead of at an invoice you did not expect.
- $2.375 per 1,000 credits
- The same rate on the smallest pack and the largest. There is no volume discount to negotiate for.
- 2,500 credits to start
- Granted on signup and good for 14 days. No card.
- Your own key, at no markup
- Bring a provider key and keep working straight past the cap. Consul takes nothing on top of it.
- 16,000credits$38
- 40,000credits$95
- 80,000credits$190
Credits stay good for twelve months and roll over month to month.
The basics
What is Consul?
One place to keep everything your company knows, with agents, apps and workflows running on top of it. You point it at the drives and tools the work already lives in, and every surface reads from that same context and writes back to it.
Who is it for?
Businesses of roughly 1 to 500 people, of any type. It is not built for one industry, and there is nothing agency-shaped about it.
Do we have to move our files into it?
No. Connections read from the tools you already use. You can also import a drive or a whole workspace if you would rather have a copy.
Credits and billing
How does pricing work?
There are no seats and no plan tiers. Work costs credits, and credits cost $2.375 per thousand at every size. Buy more from inside the workspace whenever you need them.
What do we get to start?
2,500 credits, free, good for 14 days. No card.
Do credits expire?
Purchased credits stay good for twelve months. The signup grant is the exception and expires with the trial.
Can a workspace overspend?
No. You set a cap and the workspace stops at it. Every debit is a row in an append-only ledger, so you can always answer what was spent, by whom, on which run, on which day.
Can we bring our own model key?
Yes, at no markup. Consul takes nothing on top of it.
Security
Can one workspace see another?
No. Isolation is enforced in the database under Postgres row-level security, not in application code that could forget a WHERE clause. Every scoped read and write runs through one narrow role, and there is no code path that widens it.
Do our credentials reach the model?
No. Agent work runs in a sandbox that is handed an identifier and nothing else. Provider keys, connector tokens and database credentials stay on our side of that boundary and are never written into a sandbox environment, so a prompt that talks an agent into revealing its secrets has none to reveal.
What happens to a document an agent reads?
It is treated as data, never as instructions. Untrusted content never renders on the same origin as your session and never carries your credentials.
Is our data used for training?
No.
Your company already knows most of what it needs to know
Point Consul at one drive and one tool, and see what it can answer by the end of the week.
2,500 credits to start, 14 days, no card.