Clampd sits between your agents and the tools they touch. It checks every database query, API call, file access, and shell command against policy before the call runs. If an agent goes rogue, you can kill it in under 25ms. One line to integrate, self-hosted, and your data never leaves your network.
$ pip install clampd / npm install @clampd/sdk
AI agents call tools with full credentials. One prompt injection turns your helpful assistant into an attacker with insider access.
Malicious instructions hidden in retrieved documents redirect your agent to execute harmful tool calls.
Agents inherit broad OAuth scopes. One compromised chain grants access to delete databases, send emails, transfer funds.
Injected payloads push sensitive data out through ordinary tool calls: API keys, PII, and credentials sent to an attacker's endpoint.
Multi-step agents amplify risk. A single poisoned step cascades through the entire chain with compounding damage.
Developers spin up agents with API keys. No logging, no approval, no audit trail. Your CISO doesn't know they exist.
When an agent goes rogue, there's no way to cut access in real-time. By the time you notice, the damage is done.
Clampd is a runtime gateway. Agents, IDE assistants, and MCP clients route their tool calls through it. Each call gets scored, then either passes with a scoped token or stops at the boundary.
A call is classified and policy-checked before it reaches the downstream service, then logged after. Rules evaluate in microseconds; end-to-end stays in single-digit milliseconds on commodity hardware.
Verify credentials, resolve agent profile, check kill-switch status and scoped permissions. Reject unauthenticated requests immediately.
Rules engine matches patterns against known threats. Policy engine allows or denies based on risk score, agent scope, and session context. Violations blocked in real-time.
Apply least-privilege access: each call is scoped to only the permissions it needs. Forward to the downstream tool and capture the response.
Every event is recorded with tool name, risk score, policy action, denial reason, and latency. The full audit trail. See full architecture →
Prompts are scanned for injection and leaked secrets. Tool calls are scored against 285 detection rules, scoped with Ed25519 tokens, and checked for behavioural anomalies, cross-agent correlation, and AP2 / x402 payment abuse. Rule evaluation runs in microseconds.
Every tier runs the full security pipeline — the same 285 rules, kill switch, and scanning. You upgrade for scale, team controls, and support, never to unlock detection.
The full security pipeline, free — for evaluation and small deployments.
Everything in Design Partner, at production scale for a team. Annual: $4,990 (2 months free).
Org-grade governance, identity, and support for regulated teams.
Sign up for free. No credit card required. Install the SDK and start guarding tool calls in minutes.
Enterprise? support@clampd.dev