Agent Testing Suite
Ten demos show Faramesh in real scenarios: blocking rm -rf while allowing ls, requiring approval for discounts over 30%, preventing infinite agent loops, and detecting PII in logs. Each demo targets a specific risk data loss, cost overruns, compliance and shows how policies and approvals address it. This is for evaluators and technical leads who want to see concrete outcomes before adopting.
Available demos
# | Demo | Framework | What it shows | Time |
|---|---|---|---|---|
1 | Float canonicalization | Core | 1.0, 1.00, 1 produce identical SHA-256 hashes; foundation of deterministic policy | ~3 min |
2 | LangChain delete-all prevention | LangChain | Block dangerous | ~5 min |
3 | Customer service discount control | Business | Prevent 100% discounts; require approval for 30–80% | ~5 min |
4 | AutoGen high-value approval | AutoGen | Human-in-the-loop for transactions ≥ $1,000 | ~5 min |
5 | MCP filesystem security | MCP | Path-based access control, operation restrictions | ~5 min |
6 | CrewAI infinite loop prevention | CrewAI | Rate limiting stops Agent A → B → A loops | ~4 min |
7 | Zero-trust cryptographic audit | Core | SHA-256, provenance IDs, immutable audit trail | ~4 min |
8 | Latency benchmark | Performance | Sub-2 ms overhead for local deployment | ~3 min |
9 | Healthcare PII redaction | Healthcare | Detect and redact SSN, credit cards in logs | ~4 min |
10 | DevOps security | DevOps | Allow | ~5 min |
Quick start
Interactive menu (recommended)
From the demo_agents folder:
You can run individual demos by number (1–10), run all sequentially (full), or a short quick set.
Run a single demo
Prerequisites
Start the Faramesh server (Horizon or self-hosted).
Set environment:
Install dependencies as needed:
langchain,crewai,autogen, etc.
What you’ll see
Technical: Deterministic hashing, policy-based governance at the execution gate, cryptographic audit trails, sub-2 ms latency overhead.
Business: Cost savings (e.g. prevented over-discounts), risk mitigation (blocked destructive commands), human-in-the-loop for critical decisions.
Security: Zero-trust architecture, fail-closed enforcement, rate limiting, PII detection, surgical command filtering.
Policy files for demos typically live in the server’s profiles/ or policies/ directory (e.g. langchain_filesystem_policy.yaml, customer_service_policy.yaml). See the demo script and server docs for paths.
