Hardened Runtime Guardrails for Autonomous LLM Agents
Zero-dependency protection against RFC 1918 internal network pivots, cloud instance metadata exfiltration, and unsafe AST code evaluation. Engineered for production LangChain, smolagents, and Model Context Protocol pipelines.
pip install bristlecone-guard
[00:01.04] Executing tool: VisitWebpageTool
[00:01.05] Query target: http://169.254.169.254/latest/meta-data
✖ [BLOCKED] Target Inaccessible
SSRF Protection: Refusing to resolve link-local cloud metadata IP. Network socket aborted.
[00:01.06] Audit event signed and transmitted to MCP host.
Core Primitives
Engineered for Microsecond-Scale Verification
Pre-Flight SSRF Guard
Deterministic socket address verification. Halts agent execution prior to socket creation if DNS resolves to private RFC 1918 subnets, loopback, or cloud IMDS endpoints.
AST Sandbox Inspection
Evaluates dynamically synthesized Python AST nodes. Forbids unauthorized built-ins, subshell spawns, and namespace tampering before the interpreter executes.
Deterministic JSON Sanitizer
Fast single-pass heuristic parser to restore truncated brackets, escape illegal control characters, and eliminate pipeline halting without redundant LLM retries.
Access & Licensing
Dual-Rail Infrastructure
Developer
Standard runtime security for independent builders and single-agent microservices.
- ✓ 100k guarded calls / mo
- ✓ SSRF Pre-Flight DNS Engine
- ✓ Model Context Protocol Server
- ✓ Standard API authentication
Engineering Team
Full guardrail suite for multi-agent swarms, CI/CD runners, and enterprise agents.
- ✓ 1M guarded calls / mo
- ✓ Full AST & Code Execution Guard
- ✓ Real-time audit stream & webhook alerts
- ✓ Multi-token team provisioning
Machine-to-Machine
Frictionless cryptographic micropayments for autonomous agents operating without human cards.
- ✓ Zero account or subscription required
- ✓ HTTP 402 micro-settlement
- ✓ Native on-chain validation
- ✓ Millisecond cryptographic authorization