AgentVaultThe right tools.
The right boundaries.
Bring tool registration, secret management and bounded execution into a clearer control layer.
Who may call what,
with which secret.
AgentVault is the registry where a tool becomes usable: declared scope, a named role, a credential reference and an execution boundary.
Tool registry
Every callable tool is registered with its input schema, its required permissions and the systems allowed to reach it. Unregistered tools are not callable.
Secret management
Credentials are stored and rotated as references, never as values in code or prompts. A tool receives a scoped credential for the duration of one call.
Bounded execution
Execution limits — timeout, output size, retry count — are declared per tool, so a misbehaving tool fails inside a known envelope.
Audit trails
Each call records who invoked it, under which role, with which credential reference and what the tool returned.
Try a role
against a tool.
A sample role matrix only. No secrets are stored and no tools are executed. The event number below is an illustration ID, not a cryptographic hash.
Access-policy sample
| Role | Read document | Write document |
|---|---|---|
| Reader | Allow | Deny |
| Operator | Allow | Allow |
No sample decision yet.
Illustration only. No tool was executed and no credential was accessed.
Secrets stay
out of scope here.
Two limits worth stating before you evaluate this for your own environment.
Environment review
Secret handling and access isolation must be reviewed in the target environment. This website stores no secrets and runs no agent tools.
Source and publication
Source availability and publication records do not certify the current Next revision. Review release notes, tests and deployment assumptions for your environment.
Where to go
next.
The current development branch, the running documentation, the published original and its permanent record.
The DOI identifies a published record. Consult that record for its exact version and files; it does not automatically identify the latest Next commit.
Start with your
tool surface.
Tell us which tools your agents reach for and which credentials they should never see. That is the shortest path to a useful assessment.