Skip to main content

Documentation Index

Fetch the complete documentation index at: https://aivault.moldable.sh/docs/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you through a complete flow: initialize the vault, store a secret, and invoke a capability — all without exposing the secret to calling code. Required reading: Security model (how aivault keeps secrets safe).

​
Prerequisites

  • The aivault binary installed and on your PATH (see Install)
  • On macOS/Linux, aivaultd installed alongside aivault (used by invoke by default). If you only have aivault, set AIVAULTD_DISABLE=1.
  • An API key for at least one supported provider (e.g. OpenAI)

​
1) Check vault status

aivault status
The vault auto-initializes on first run with safe defaults:
  • macOS: uses the system Keychain
  • Other platforms: uses the file provider with a key at ~/.aivault/keys/kek.key (outside the vault directory)
If you prefer a passphrase-protected vault (manual unlock after restart), initialize explicitly: 
aivault init --provider passphrase --passphrase "your-passphrase"
See Vault lifecycle for all provider options.

​
2) Store a secret

aivault secrets create \
  --name OPENAI_API_KEY \
  --value "sk-..." \
  --scope global
Because OPENAI_API_KEY matches the built-in registry, this automatically:
  • Pins the secret to the openai provider (it can only be used for OpenAI hosts)
  • Provisions the openai credential
  • Enables all 17 OpenAI capabilities (chat, transcription, embeddings, images, etc.)

​
3) Browse capabilities

# List all capabilities and their readiness
aivault capability list

# Inspect a specific capability
aivault capability describe openai/chat-completions
The describe command shows allowed methods, path prefixes, and example invocations.

​
4) Invoke a capability

aivault invoke openai/chat-completions \
  --method POST \
  --body '{"model":"gpt-5.2","messages":[{"role":"user","content":"hello"}]}'
The broker validates the request against capability policy, decrypts the secret from the vault, injects the auth header, and proxies the request to api.openai.com. The response is returned with auth-class headers stripped. For structured output: 
# JSON output
aivault json openai/chat-completions \
  --method POST \
  --body '{"model":"gpt-5.2","messages":[{"role":"user","content":"hello"}]}'

# Markdown output
aivault markdown openai/chat-completions \
  --method POST \
  --body '{"model":"gpt-5.2","messages":[{"role":"user","content":"hello"}]}'

​
5) Verify security posture

# View audit log
aivault audit

# Confirm secrets are never printed
aivault secrets list
Secret values are never printed by any command. The audit log records every create, rotate, and invoke event.

​
What just happened

You (CLI)
  │
  │  "invoke openai/chat-completions with this body"
  ▼
Broker runtime
  ├─ Validated: POST is allowed, path matches /v1/chat/completions
  ├─ Resolved: openai credential → vault secret OPENAI_API_KEY
  ├─ Decrypted: secret from vault (XChaCha20-Poly1305)
  ├─ Injected: Authorization: Bearer sk-... into outgoing request
  ├─ Host: api.openai.com (derived from capability, not from caller)
  ▼
api.openai.com
  │
  ▼
Response returned to you (auth headers stripped)
The calling code (you, a script, an agent) never saw the secret. Even if the calling code were compromised, it could only make requests the capability policy allows — to the hosts the registry defines.

​
Multi-secret providers

Some providers require multiple secrets (e.g. Trello needs both an API key and a token). The credential auto-provisions once all required secrets are present: 
aivault secrets create --name TRELLO_API_KEY --value "your-key" --scope global
# → Waiting for TRELLO_TOKEN to complete trello credential

aivault secrets create --name TRELLO_TOKEN --value "your-token" --scope global
# → Credential auto-provisioned: trello (17 capabilities enabled)

​
Bulk import

If you have multiple secrets to store: 
aivault secrets import \
  --entry OPENAI_API_KEY=sk-... \
  --entry ANTHROPIC_API_KEY=sk-ant-... \
  --entry GITHUB_TOKEN=ghp-... \
  --scope global

​
Isolating test data

To use an isolated vault for testing without touching your real secrets: 
export AIVAULT_DIR="$(mktemp -d)"
aivault status
aivault secrets create --name OPENAI_API_KEY --value sk-test --scope global
aivault invoke openai/chat-completions --body '...'
Next: Security model If you’re setting this up for an agent-heavy environment, also see: