> ## Documentation Index
> Fetch the complete documentation index at: https://infisical.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Agent Vault
> Run AI agents with the access, context, and capabilities they need to do work.
**Infisical Agent Vault** equips AI agents with what they need to do work, starting with secure access to the services they call: LLM providers, GitHub, Slack, and more.
Agents like Claude Code or OpenClaw work as they always do. Their requests leave your network with the real credential attached, and the agent never holds it.
```mermaid theme={"dark"}
flowchart LR
subgraph private["Your private network"]
agent["AI agent"]
proxy["Agent Vault proxy"]
end
subgraph internet["Public internet"]
apis["api.slack.com
api.github.com
..."]
end
infisical["Infisical"]
agent -->|"request with a session token"| proxy
proxy -->|"request with the credential"| apis
proxy -->|"fetches the credentials this session may use"| infisical
```
## Why Agent Vault
Traditional secrets management involves returning credentials back to applications and services.
This isn't suitable for AI agents because they're vulnerable to credential exfiltration via [prompt injection](https://en.wikipedia.org/wiki/Prompt_injection); an attacker could craft a malicious prompt or payload and exfiltrate credentials from an agent back to the attacker.
Enter Infisical Agent Vault, which prevents credential exfiltration by brokering access at the network boundary. The agent gets an access session that works only for the services you allowed, and only until it expires or you revoke it.
## How it works
You describe what an agent may use in an access bundle: one service per API the agent may reach, each with the credential the proxy attaches. You grant the access bundle to whoever runs agents, and they create a session with it to launch the agent. Requests to those services get the right credential on the way out, and everything else is untouched. The agent needs no code changes.
## Get started
Create an access bundle with a service for GitHub, enroll a proxy, and launch an agent that makes authenticated calls with a token it never sees.
## Learn more
Services, credential types, host patterns, and who can run agents with an access bundle.
Creating one, how long it lives, revoking it, and what a running agent can reach.
Where the proxy runs, enrolling it, traffic policy, and certificate trust.
Every flag of `infisical agent-vault proxy` and `infisical agent-vault run`.