Secure Identities
for AI Agents

Give your agents credentials, phone numbers, emails, and API keys. Everything stored in a vault architecture trusted by enterprises for sensitive PII.

from footprint import agent_identity
# Load an identity vault
identity_vault = agent_identity.get(name="review_analyst")
# Generate a scoped token for the agent
token = agent_identity.generate_token(
    expiry="...",
    scopes=["vault:read", "identity:login", "..."]
)
# Pass the token to the agent
run_my_agent(token)

Define identities, generate scoped tokens, and pass them to your agents.

Core Platform

Everything agents need to operate

A complete identity layer purpose-built for autonomous agents. Secure by default, observable by design.

Identity Profiles

Define complete identities for each agent — credentials, phone numbers, emails, and API keys in one place.

Vault Architecture

Sensitive data encrypted at rest and in transit. The same vault that enterprises trust for PII, now for your agents.

MCP Server

Built-in Model Context Protocol server. Agents can read, write, and manage their identity through a standardized interface.

SDKs & Libraries

First-class Python and TypeScript SDKs. Install, import, and integrate agent identity in minutes.

Access Control

Granular permissions define exactly what each agent can access. Scope credentials, rotate keys, and revoke instantly.

Vault Memory

Persistent, mutable memory per agent. Every interaction can read from and write to a secure context store.

Agentic Identity

Give your agents
a secure identity

  • Full browser-based authentication — works with all browser automation frameworks.

  • Expose tools via SDK or MCP to let your agents discover who they are and what they can do.

  • Read and write sensitive data without dumping it into the context window.

login.py
# Authenticate into a live browser session
agent_identity.login(browser_session)

Vault Memory

Persistent context
across every run

Each agent gets a mutable memory store backed by the vault. Read and write context between sessions — preferences, state, tool outputs, conversation summaries — all encrypted and scoped.

  • •Read/write memory from any SDK or MCP
  • •Encrypted at rest with per-agent keys
  • •Versioned with full history and rollback
vault-memory.py
# Write to vault memory
agent_identity.memory.write("customer_communication_update", {
    "customer": "john.doe@example.com",
    "sent": "user@acme.com",
    "status": "delivered",
    "summary": "The customer was notified we need to collect their bank statement for verification."
})
# Read it back agentically
agent_identity.memory.read(filters=["john.doe@example.com", "verification email"])

Integrate Anywhere

SDKs, MCP, and CLI

First-class support for the tools your agents already use.

Py

Python SDK

pip install footprint-agent

Async-first, type-safe Python client. Works with LangChain, CrewAI, and any agent framework.

TS

TypeScript SDK

npm i @footprint/agent

Full TypeScript types, tree-shakeable, works in Node.js, Deno, Bun and edge runtimes.

MCP Server

footprint mcp start

Drop-in MCP server for Claude, Cursor, and any MCP-compatible client. Zero config.

Auditability

Every access
logged and traceable

Complete audit trail for every credential read, memory write, and identity access. Know exactly which agent accessed what, when, and whether it was allowed.

<1ms
Log latency
100%
Coverage
90d
Retention
Access LogLive
12:04:31support_botcredential.readstripe_api_keyallowed
12:04:29support_botmemory.writelast_ticket_idallowed
12:03:58data_agentcredential.readstripe_api_keydenied
12:03:45deploy_botcredential.readaws_access_keyallowed
12:02:12support_botidentity.reademailallowed

Give your agents
a secure identity

Start building with the Agent Identity Vault. Free during beta.