Argus + tokenecon — playground
Two open-source tools for running AI agents in production, by Karmendra Pandey. Argus is a policy gateway that checks every prompt and reply before it reaches a model. tokenecon prices agent work before it runs, so model choice becomes a budget decision. Both panels below run entirely in your browser — no API keys, no network calls, and nothing you paste is stored.
This page is a faithful JavaScript port of the demo’s detector patterns, policy behavior, and dated pricing snapshot from the Python packages — it does not execute the Python packages themselves. See the README for details.
Argus check
Argus sits in front of an LLM and enforces a security policy on every request and response. Paste a prompt or an agent message and run it through the same detectors and policy behavior as the gateway (argus-gateway 0.1.2).
Profile used in this demo: Demo (corrected rule set) — secrets block · SSNs and card numbers block · other PII is highlighted (flagged, not removed) · prompt-injection patterns are flagged. In the gateway, a profile is assigned per API key and the same checks also run on model replies.
GitHub: karmendra8386/argus · Cite: doi.org/10.5281/zenodo.23427027
Try these:
tokenecon cost estimate
tokenecon decomposes agent spend and prices it per model, so a team can forecast cost before deployment instead of reading about it on the invoice. Pick a model and token counts — or estimate the input tokens from the text in Panel 1 — and see the same workload priced on every model in the snapshot.
GitHub: karmendra8386/agentic-ai-token-economics · Cite: doi.org/10.5281/zenodo.23427025