wardn

Credential isolation for AI agents. Agents never see real API keys — structural guarantee, not policy.

The Problem

Every AI agent framework today stores API keys in environment variables or .env files. A compromised agent, malicious skill, or commodity stealer gets full access to your credentials.

~/.env              → OPENAI_KEY=sk-proj-real-key      # plaintext, readable by anyone
agent context       → "Use OPENAI_KEY=sk-proj-real-key" # leaked into LLM context window
agent logs          → Authorization: Bearer sk-proj-... # sitting in log files

The Fix

Wardn vaults credentials with AES-256-GCM encryption and gives agents useless placeholder tokens. Real keys are injected at the network layer — agents never touch them.

agent environment   → OPENAI_KEY=wdn_placeholder_a1b2c3d4e5f6g7h8   (useless)
wardn vault         → OPENAI_KEY=sk-proj-real-key                     (encrypted)
agent logs          → Authorization: Bearer wdn_placeholder_a1b2...   (useless)
LLM context window  → wdn_placeholder_a1b2c3d4e5f6g7h8               (useless)

Architecture

flowchart TB
    subgraph Agent["AI Agent Process"]
        A1["Agent Code"]
        A2["ENV: OPENAI_KEY=wdn_placeholder_a1b2..."]
    end

    subgraph Wardn["wardn daemon · localhost:7777"]
        direction TB
        P["HTTP Proxy"]
        MCP["MCP Server\n(stdio)"]

        subgraph Pipeline["Request Pipeline"]
            direction LR
            S1["Identify\nAgent"] --> S2["Resolve\nPlaceholder"] --> S3["Check\nAuth"] --> S4["Rate\nLimit"] --> S5["Inject\nReal Key"]
        end

        subgraph ResponsePipeline["Response Pipeline"]
            direction RL
            R1["Strip Real\nKeys"] --> R2["Replace with\nPlaceholders"]
        end

        subgraph Vault["Encrypted Vault"]
            V1["AES-256-GCM"]
            V2["Argon2id KDF"]
            V3["Placeholder Map\nper agent × credential"]
        end
    end

    subgraph External["External APIs"]
        E1["api.openai.com"]
        E2["api.anthropic.com"]
        E3["..."]
    end

    A1 -- "placeholder token\nin headers/body" --> P
    A1 -. "MCP: get_credential_ref\nlist_credentials\ncheck_rate_limit" .-> MCP
    MCP -. "placeholder token\n(never real keys)" .-> A1
    P --> Pipeline
    Pipeline --> External
    External --> ResponsePipeline
    ResponsePipeline -- "response with\nplaceholders only" --> A1
    Pipeline <--> Vault
    ResponsePipeline <--> Vault

    style Agent fill:#1a1a2e,stroke:#e94560,color:#fff
    style Wardn fill:#0f3460,stroke:#16213e,color:#fff
    style Pipeline fill:#16213e,stroke:#e94560,color:#fff
    style ResponsePipeline fill:#16213e,stroke:#e94560,color:#fff
    style Vault fill:#1a1a2e,stroke:#00d2ff,color:#fff
    style External fill:#0a0a0a,stroke:#533483,color:#fff

How It Works

Agent sends request with placeholder in Authorization header
         │
         ▼
┌─────────────────────────┐
│      wardn proxy        │
│    localhost:7777        │
│                         │
│  1. Identify agent      │
│  2. Resolve placeholder │
│  3. Check authorization │
│  4. Check rate limit    │
│  5. Inject real key     │
│  6. Forward request     │
│  7. Strip key from resp │
│  8. Return to agent     │
└─────────────────────────┘
         │
         ▼
   External API (only place real key exists in transit)

Demo

Install

cargo install wardn

Quick Start

# Create an encrypted vault and store your keys
wardn vault create
wardn vault set OPENAI_KEY
wardn vault set ANTHROPIC_KEY

# Set up Claude Code integration (one command)
wardn setup claude-code

That's it. Claude Code now uses wardn's MCP server to get placeholder tokens instead of reading real keys from your environment.

What happens next

1. Claude Code calls get_credential_ref → gets wdn_placeholder_a1b2... (not the real key)
2. Agent sends request with placeholder through wardn proxy
3. Proxy swaps placeholder for real key, forwards to API
4. Proxy strips real key from response before returning to agent

The real key never enters the agent's memory, logs, or LLM context window.

Manual setup

# Get a placeholder token (never the real key)
wardn vault get OPENAI_KEY
# → wdn_placeholder_a1b2c3d4e5f6g7h8

# List stored credentials (names only, no values)
wardn vault list

# Start the proxy
wardn serve

# Start proxy + MCP server for Claude Code / Cursor
wardn serve --mcp --agent my-agent

CLI Reference

Vault Management

wardn vault create                        # create encrypted vault
wardn vault set OPENAI_KEY                # store credential (prompts for value, no echo)
wardn vault get OPENAI_KEY                # get placeholder token (never the real value)
wardn vault get OPENAI_KEY --agent bot    # get placeholder for specific agent
wardn vault list                          # list all credentials
wardn vault rotate OPENAI_KEY             # rotate value, placeholders unchanged
wardn vault remove OPENAI_KEY             # remove credential

# Custom vault path
wardn --vault /path/to/vault.enc vault list

Proxy Server

wardn serve                               # HTTP proxy on 127.0.0.1:7777
wardn serve --host 0.0.0.0 --port 8080    # custom bind address
wardn serve --config wardn.toml           # load config with rate limits + ACLs
wardn serve --mcp --agent my-agent        # proxy + MCP server (stdio)

Claude Code / Cursor Integration

wardn setup claude-code                   # register wardn as MCP server in Claude Code
wardn setup cursor                        # register wardn as MCP server in Cursor

# Or manually:
claude mcp add --transport stdio --scope user wardn -- wardn serve --mcp --agent claude-code

What wardn setup does

1. Prompts for your vault passphrase and verifies it can open the vault
2. Finds the wardn binary path on your system
3. Registers wardn as an MCP server:
   • Claude Code: runs claude mcp add with WARDN_PASSPHRASE in the env config
   • Cursor: writes to ~/.cursor/mcp.json with the passphrase in env
4. On next launch, the IDE spawns wardn serve --mcp as a subprocess

Verifying it works

After running setup, restart your IDE and try these prompts:

"List my wardn credentials"
→ Claude calls list_credentials, shows credential names (never values)

"Get me a reference to OPENAI_KEY"
→ Claude calls get_credential_ref, gets wdn_placeholder_... (not the real key)

"Check my rate limit for OPENAI_KEY"
→ Claude calls check_rate_limit, shows remaining quota

MCP tools available

Credential Migration

wardn migrate --dry-run                             # audit Claude Code dir for exposed keys
wardn migrate --source claude-code                  # scan + migrate to vault
wardn migrate --source open-claw                    # scan OpenClaw config
wardn migrate --source directory --path ./my-proj   # scan any directory

Automation

For CI/scripts, set WARDN_PASSPHRASE and WARDN_VALUE env vars to skip interactive prompts:

WARDN_PASSPHRASE=my-pass wardn vault list
WARDN_PASSPHRASE=my-pass WARDN_VALUE=sk-proj-xxx wardn vault set OPENAI_KEY

Library API

Add to your Cargo.toml:

[dependencies]
wardn = "0.3"

Vault Operations

use wardn::{Vault, config::CredentialConfig};

// Create an encrypted vault
let vault = Vault::create("vault.enc", "my-passphrase")?;

// Store a credential
vault.set_with_config("OPENAI_KEY", "sk-proj-real-key-123", &CredentialConfig {
    allowed_agents: vec!["researcher".into(), "writer".into()],
    allowed_domains: vec!["api.openai.com".into()],
    rate_limit: Some(RateLimitConfig { max_calls: 200, per: TimePeriod::Hour }),
})?;

// Agent gets a placeholder (not the real key)
let placeholder = vault.get_placeholder("OPENAI_KEY", "researcher")?;
// → "wdn_placeholder_a1b2c3d4e5f6g7h8"

// Rotate the real key — all placeholders keep working
vault.rotate("OPENAI_KEY", "sk-proj-new-key-456")?;

HTTP Proxy

use wardn::daemon::{Daemon, DaemonConfig};

let daemon = Daemon::new(vault, DaemonConfig::default());
daemon.serve_proxy().await?;

MCP Server

use wardn::mcp::WardenMcpServer;

// Serve over stdio (for Claude Code, Cursor, etc.)
WardenMcpServer::serve_stdio(vault, rate_limiter, "agent-id".into()).await?;

MCP tools exposed (read-only, no credential values ever returned):

Security Properties

What This Defeats

How Is This Different From...

Trust Boundary

Wardn concentrates trust in a single local process (the proxy) instead of spreading it across every plugin, tool, and LLM context window. This is a smaller attack surface, not zero attack surface:

• The proxy runs locally as a subprocess spawned by your IDE or shell — same trust level as your kernel
• The vault is encrypted at rest and only decrypted in-memory with your passphrase
• If your local machine is fully compromised, wardn can't help (nothing can)
• The placeholder token is a bearer token to the proxy — but it only works via localhost:7777, not against real APIs, and can be rate-limited and revoked per-agent

Audit Logging

Every credential access is logged with a unique request ID for traceability:

INFO request_id=a1b2c3 agent=claude-code method=POST domain=api.openai.com path=/v1/chat/completions proxy request received
INFO request_id=a1b2c3 agent=claude-code credential=OPENAI_KEY domain=api.openai.com credential injected
INFO request_id=a1b2c3 agent=claude-code upstream_status=200 credentials_injected=1 credentials_stripped=0 proxy request completed

Set RUST_LOG=wardn=info (or debug/trace) to control verbosity. Logs go to stderr, never stdout.

Configuration

[warden]
vault_path = "~/.vibeguard/vault.enc"

[warden.credentials.OPENAI_KEY]
rate_limit = { max_calls = 200, per = "hour" }
allowed_agents = ["researcher", "writer"]
allowed_domains = ["api.openai.com"]

[warden.credentials.ANTHROPIC_KEY]
rate_limit = { max_calls = 100, per = "hour" }
allowed_agents = ["researcher"]
allowed_domains = ["api.anthropic.com"]

Project Structure

wardn/
├── src/
│   ├── main.rs             # CLI entry point (clap + tokio)
│   ├── cli/
│   │   ├── mod.rs          # Clap argument definitions
│   │   ├── vault_cmd.rs    # Vault subcommand handlers
│   │   ├── serve_cmd.rs    # Serve subcommand handler
│   │   ├── setup_cmd.rs    # Claude Code / Cursor MCP setup
│   │   └── migrate_cmd.rs  # Migrate subcommand handler
│   ├── lib.rs              # Public API, WardenError
│   ├── config.rs           # TOML configuration parsing
│   ├── vault/
│   │   ├── mod.rs          # Vault CRUD operations
│   │   ├── encryption.rs   # AES-256-GCM + Argon2id + zeroize types
│   │   ├── storage.rs      # On-disk format (WDNV), atomic writes
│   │   └── placeholder.rs  # Token generation, per-agent isolation
│   ├── proxy/
│   │   ├── mod.rs          # HTTP proxy server (axum)
│   │   ├── inject.rs       # Credential injection into requests
│   │   ├── strip.rs        # Credential stripping from responses
│   │   └── rate_limit.rs   # Token bucket rate limiter
│   ├── mcp/
│   │   ├── mod.rs          # MCP server (rmcp, stdio transport)
│   │   └── tools.rs        # Tool parameter/response types
│   ├── migrate/
│   │   ├── mod.rs          # Migration orchestrator + risk scoring
│   │   └── scanners/
│   │       └── credentials.rs  # API key pattern scanner
│   └── daemon/
│       └── mod.rs          # Daemon (proxy + MCP in single process)
└── tests/
    ├── cli_tests.rs        # CLI integration tests
    ├── vault_tests.rs      # Vault integration tests
    └── proxy_tests.rs      # Proxy integration tests

Vault Encryption

flowchart LR
    subgraph Input
        Pass["Passphrase"]
        Salt["Random Salt\n(16 bytes)"]
        Creds["Credentials\n(JSON)"]
    end

    subgraph KDF["Key Derivation"]
        Argon["Argon2id\nm=19456 t=2 p=1"]
    end

    subgraph Encrypt["Encryption"]
        AES["AES-256-GCM"]
        Nonce["Random Nonce\n(12 bytes)"]
    end

    subgraph Output["WDNV File"]
        direction TB
        Magic["WDNV (4B)"]
        Ver["Version (2B)"]
        SaltOut["Salt (16B)"]
        Payload["Nonce ‖ Ciphertext ‖ Tag"]
    end

    Pass --> Argon
    Salt --> Argon
    Argon -- "256-bit key" --> AES
    Creds --> AES
    Nonce --> AES
    AES --> Payload

    style Input fill:#1a1a2e,stroke:#e94560,color:#fff
    style KDF fill:#16213e,stroke:#00d2ff,color:#fff
    style Encrypt fill:#16213e,stroke:#00d2ff,color:#fff
    style Output fill:#0f3460,stroke:#533483,color:#fff

File Format

Bytes 0-3:   Magic "WDNV"
Bytes 4-5:   Version (u16 LE)
Bytes 6-21:  Argon2id salt (16 bytes)
Bytes 22+:   AES-256-GCM encrypted payload (nonce ‖ ciphertext ‖ tag)

Part of VibeGuard

Wardn is the credential isolation layer of VibeGuard — a security daemon for AI agents. Other planned modules:

• Sentinel — prompt injection firewall
• CloakPipe — PII redaction middleware
• Watcher — audit log + dashboard

License

MIT