For CISOs and security teams
How many non-human identities do we have now, and what can each one reach?
Every agent is a new identity touching your systems. LeafCraft gives each one its own login with its role's access, keeps keys in a vault and fences in outside agents.
What's changing for you
Every agent is a non-human identity
Most of them act under a shared login, with the access of whoever set them up.
Outside agents are an attack surface
A reply from an outside agent that is read as an instruction is prompt injection.
Keys end up in config
Agents built in sprints keep keys where sprints keep them.
Two risks the layer removes
The agents you can't see, and the model that gets retired.
Agents you don't know about
In your company, agents arrive through vendor updates and team pilots. LeafCraft finds all of them, gives each an owner, and brings it under your rules or removes it.
A model that gets retired
An agent built straight on one model has to be rebuilt and re-approved when that model is retired or repriced. On LeafCraft you switch the model for that agent in config and keep everything else. Months and a budget, saved.
- The model is retired or repriced
- Prompts, tests and sign-offs rebuilt
- Months of work, a new approval cycle
- The history stays with the old tool
- Pick another model for that agent
- Switching is a config change
- Test it against your own record first
- Rules, limits, owner and memory stay
What keeps you up
Shared and borrowed access
Agents often run on the login of the person who set them up. They get far more access than their job needs.
Instructions hidden in replies
Replies from outside agents and tools can carry instructions. They must be read as data, not orders.
Keys spread across vendors
Every new console can become one more place holding keys to your clouds and systems.
Why it matters now
One layer under every agent and every person.
You need to know how many non-human identities you have, what each one can reach, and that none of them can be turned against you. The layer gives every agent its own login and only its role's access, fences in outside agents, reads their replies as data, keeps keys in a vault, and stops everything with one switch.
What LeafCraft gives you
Its own login, its role's access
Every agent gets its own identity, never a shared one. It uses only the tools, systems and actions you list, and anything else is refused before it runs.
Keys stay in a vault
Use the built-in encrypted vault or one of 14 you already run. LeafCraft will not start if a key is typed into the config, and no screen shows a key back.
Outside agents fenced in
They reach only the addresses you allow and get only the fields you choose. Agents that keep failing are cut off.
Guardrails that stay on
Signed webhooks carry a timestamp and a one-time code, so a captured request cannot be replayed.
What changes for you
Ask any AI agent vendor
Five questions for your next vendor meeting. They are the ones your auditor and your board will ask you. Ask them first.
- Does your agent get its own login, or does it act with a person's access?
- Exactly which systems, actions and data can it reach, and who can widen that?
- Where are the keys it uses kept, and who on your side can see them?
- Can you replay what the agent did on our systems, step by step?
- Which of our data leaves our environment, and where does it go?
How it helps your people
Where your peers start.
Industries
Questions teams ask
Does code from agents run with network access?
Not unless you allow it. Code runs in a sandbox with no access to keys and no network by default.
Are tenants kept apart?
Yes. Each tenant's sessions, memory and events are stored separately from every other tenant's.
Can we pause everything during an incident?
Yes. One switch pauses the whole operation across every server and cancels work in progress. It holds through a restart.
Bring your hardest question.
We'll answer it on one of your own processes, with the record to show for it.