Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@AgentVault MCP Server get my Stripe API key from the prod vault"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
The official open-source client for AgentVault — the API-first password/secrets manager for AI agents. An agent gets a real vault, stores secrets encrypted at rest, and fetches exactly one credential at runtime via an av://vault/secret/field reference under a scoped, revocable token.
This package is MIT-licensed. The AgentVault server (vaults, AES-256-GCM encryption, scoped-token issuance, audit) is a Vektra Industries product.
pip install agentvault-sdk # or: uv pip install agentvault-sdkRelated MCP server: Kova Mind MCP Server
from agentvault_sdk import AgentVaultClient
av = AgentVaultClient("https://vault.example.com", token="av_...")
av.create_vault("prod")
av.put_secret("prod", "stripe", {"api_key": "sk_live_..."})
# fetch one secret at runtime — no hardcoded keys
key = av.resolve("av://prod/stripe/api_key")Expose AgentVault tools to any MCP-capable agent over stdio:
export AGENTVAULT_URL=https://vault.example.com
export AGENTVAULT_TOKEN=av_...
agentvault-mcpTools: av_create_vault, av_put_secret, av_get_secret, av_resolve, av_list_secrets, av_issue_token, av_rotate_secret.
© 2026 Pablo Navarro / Vektra Industries. MIT.
Create a new vault.
Inputs: name (required).
Get the decrypted field map of a secret.
Inputs: name (required), vault (required).
Mint a scoped Agent Access Token (requires admin token).
Inputs: name, vaults, secrets, permissions.
List secret names in a vault.
Inputs: vault (required).
Create or update a secret (name + field map) in a vault.
Inputs: name (required), vault (required), fields (required).
Resolve an av://vault/secret/field reference to a single value.
Inputs: reference (required).
Rotate (replace) a secret's value.
Inputs: name (required), vault (required), fields (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 7 tools
While av_get_secret and av_resolve both retrieve secret data, av_get_secret returns the full field map while av_resolve targets a specific field via reference. av_put_secret and av_rotate_secret are similarly distinct (create/update vs. replace). The purposes are mostly clear, though the boundary between get/resolve and put/rotate could momentarily confuse.
The av_ prefix is consistent, and most tools follow av_<verb>_<noun> (create_vault, put_secret, get_secret, list_secrets, issue_token, rotate_secret). The verb style is uniform, but 'av_resolve' lacks an explicit object, a minor deviation.
With 7 tools, the set is well-scoped for a vault management server, fitting the typical 3-15 range. Each tool addresses a distinct operation without redundancy or bloat.
The server covers the core lifecycle of vaults and secrets: create, read, update, list, rotate, and token issuance. However, there is no delete operation for vaults or secrets, which could be considered a gap for complete lifecycle management, though it may be intentional for security.
Shared long-term memory vault for AI agents with 20 MCP tools.
Secure and manage MCP servers through the Speakeasy AI Control Plane from your agent.
A secret store for AI agents: the agent never sees the plaintext.
Shared memory for your AI tools: tasks, notes and context every tool can read and write over MCP.
