Product changelog
Every significant update to sipi.bot — from the founding $12,400 bill to today's incident database.
AI Agent Incident Database launched
Open, sourced database of 34 AI-agent incidents — CC BY 4.0, JSON/CSV/JSONL, live at /incidents/. First canonical source for agent-spend loss data.
Spend-policy generator + risk calculator
Two free interactive tools: the Agent Spend Risk Calculator (score 1–10 from your agent config) and the Spend Policy Generator (JSON, YAML, curl output). Live at /tools/.
pSEO expansion: scenarios, redflags, calculators, guides
Added 4 new content sections — /scenarios/, /redflags/, /calculators/, /guides/ — covering specific agent-risk scenarios and how each rule type addresses them.
MCP server launched
sipi.bot is now a native MCP server. Any agent framework that speaks MCP (Claude Code, Cursor, LangChain) can use the firewall as a standard tool with no SDK.
Protected-by badge goes live
Free embeddable image badge showing live firewall stats. No JS, no account needed. Embed on any README, docs site, or landing page.
Sipi.bot eval gym: 53/53
The evaluation harness now covers all 53 scenarios across all 6 rule types. Results published as machine-readable JSON at /eval and human-readable at /eval-report/.
Sipi.bot v1 launched
First hosted deployment: Team $99/mo, Business $499/mo, free self-host (MIT). Core engine: one call → APPROVED/BLOCKED/FLAGGED in <5ms.
Sipi.bot founded
After a $12,400 overnight Azure bill from a runaway agent loop, the belief that every autonomous agent needs a deterministic spend firewall became a company.
What's next
We're building the canonical reference for AI agent spend governance. The database grows weekly. Follow the blog for deep dives and the incident database for the underlying data.