Live on Railway — crawdad-production.up.railway.app

OpenClaw has no security layer. Now it does.

CrowdStrike and Cisco flagged OpenClaw's security gaps — prompt injection, data exfiltration, unvetted skills, no access control. Crawdad is the open-source runtime trust layer that fixes all of them. Works with OpenClaw, LangChain, CrewAI, AutoGen, and any agent framework.

Read the Guide → View on GitHub
490+Tests passing
70+API endpoints
10Rust crates
0Unsafe code
terminal
# Secure your OpenClaw agent in 30 seconds
pip install crawdad-sdk

# Or hit the API directly
curl -X POST https://crawdad-production.up.railway.app/agents \
  -H "X-API-Key: $KEY" \
  -d '{"display_name": "my-agent"}'

AI agents are running without guardrails

OpenClaw has 68K+ GitHub stars and runs on thousands of machines. CrowdStrike, Cisco, and security researchers have all flagged the same gaps:

Prompt Injection

Malicious instructions embedded in data can hijack the agent's capabilities. No semantic firewall exists to detect or block them.

Data Exfiltration

Cisco found third-party skills that exfiltrated data without user awareness. No content filtering or PII detection exists.

Unvetted Skills

Skills are directories with a markdown file. No hash verification, no attestation, no capability policy enforcement.

No Access Control

Agents run with the user's full permissions. No identity verification, no policy engine, no action authorization.

Sources: CrowdStrike, Cisco, Microsoft Security, Giskard, Penligent — documented across multiple security advisories and CVE-2026-25253.

Seven pillars of agent trust

Crawdad wraps your OpenClaw agent (or any agent) in a complete security layer. Every pillar addresses a real vulnerability that exists today.

🔑

Agent Identity

Ed25519 keypairs, DID documents, encrypted credential vaults, three-level kill switch, and purpose-bound scoped tokens with automatic expiry.

8 endpoints
🛡️

Semantic Firewall

Structural deobfuscation, 27 injection patterns, output guard with exfiltration detection, and instruction density scoring for slow escalation attacks.

4 endpoints
⚖️

Policy Engine

5-factor risk scoring, Rule of Two enforcement, behavioral baselines with anomaly detection, and configurable permit/escalate/deny decisions.

6 endpoints
🧠

Memory Integrity

Merkle-chained entries with Ed25519 signatures, firewall-gated writes, surgical rollback, compaction seals, and write anomaly detection.

7 endpoints
🔧

Skill Attestation

SHA-256 manifest verification, static analysis, capability policies, version validation, typosquat detection, runtime monitoring, and SBOM analysis.

11 endpoints
📡

Comms Governance

Signed envelopes, delegation chains with scope reduction, content filtering, collusion detection, cascade breakers, quarantine zones, and provenance tracking.

13 endpoints
🔒

Privacy & Compliance

15-category PII detection, 4 transform modes, consent management, DSAR engine, 10-jurisdiction compliance, differential privacy, and reporting.

14 endpoints

Built for production agent deployments

Beyond the core pillars, Crawdad includes hardened security features designed for real-world multi-agent systems.

R2

Rule of Two

No agent holds untrusted input + sensitive data + code execution simultaneously. Auto-deny on violation.

CB

Cascade Breaker

Per-agent fan-out limits with auto-trip on >50% error rate. Three isolation levels: Soft, Hard, Quarantine.

PP

Provenance Propagation

Trust decay (10%/hop, floor 0.1) prevents trust laundering where untrusted data gains trust through intermediaries.

DC

Delegation Chains

Validates multi-hop delegations for monotonic scope reduction, depth limits, and circular prevention.

CS

Compaction Seals

Cryptographic seals over compacted memory with SHA-256 hashes, composite provenance, and tamper-evident verification.

SB

SBOM Analysis

CycloneDX bills of materials with semver vulnerability matching, license policy enforcement, and risk scoring.

DP

Differential Privacy

Laplace and Gaussian noise mechanisms for aggregate queries with privacy budget tracking and exhaustion enforcement.

MT

Multi-Tenancy

Per-tenant data isolation, scoped API keys, admin management endpoints, and configurable agent quotas.

Secure your OpenClaw agent in 5 lines

pip install crawdad-sdk[openclaw]. The Python SDK wraps all 70+ endpoints with type hints, error handling, and context manager support. The OpenClaw middleware intercepts inbound, outbound, and tool execution.

from crawdad.openclaw import CrawdadMiddleware

mw = CrawdadMiddleware(
  base_url="https://crawdad-production.up.railway.app",
  api_key="your-key"
)

# Scan inbound messages for prompt injection
result = mw.scan_inbound("user message")

# Gate tool execution through policy
result = mw.authorize_action(agent_id, "shell_exec", "/bin/bash")

# Scan outbound for PII and credentials
result = mw.scan_outbound("Contact john@example.com")
safe = result["redacted"]

Everything you need

OpenClaw Integration
Middleware, skill scanner, memory guard, policy gate, credential detector. Covers all 10 documented vulnerability categories.
Python SDK on PyPI
48 methods, 35+ Pydantic models, typed exceptions. pip install crawdad-sdk[openclaw].
OpenAPI 3.1 Spec
62 operations, 113 schemas. Import into Postman, Swagger UI, or generate clients in any language.
CLI Onboarding
crawdad openclaw init / scan / audit / protect. Zero-friction setup for OpenClaw installations.

Run anywhere

Single binary, SQLite persistence, zero external dependencies. Deploy in seconds.

Docker

Pull and run. Persistent volume for SQLite. Done.

docker run -p 3000:3000 crawdad

Kubernetes

Production Helm chart with HPA, secrets, PVC, and ingress.

helm install crawdad ./helm/crawdad

Railway

One-click deploy. Already running in production.

railway up

From Source

Clone, build, run. Rust 1.75+, that's it.

cargo run --release

Secure your agents today

Built for OpenClaw, works with any agent framework. Open source. BSL 1.1 licensed. pip install crawdad-sdk[openclaw].