AI agent security runtime (CLI) · In development · 2026
SVTPrivateAI
The model proposes. The runtime decides.
- Type
- AI agent security runtime (CLI)
- Status
- In development
- Year
- 2026
- Disciplines
- AI infrastructure
Security engineering
Developer tool - Technology
- Python
Pydantic
Typer
Rich
In development. Version 0.1 (alpha): the foundation is complete and tested; not yet hardened for multi-tenant or internet-facing use.
Overview
The project in brief
A local-first Python runtime that runs AI agents under scoped capabilities, a deterministic policy engine, human approvals and a hash-chained audit log.
Agents are declared in YAML with scope-aware capabilities — read wide, write only where needed — and every tool call goes through one controlled path: capability check, policy evaluation, an optional human approval bound to that exact call, then a tamper-evident audit record that names the rule that acted.
Providers include Anthropic, OpenAI-compatible endpoints and a deterministic mock used by default, so it runs without an API key. MCP servers are treated as untrusted extensions, and secrets are resolved inside tools, never placed in model context. The svt CLI checks, explains and audits every decision.
- Built by
- SIMONVILSTECH
- Nature
- Own product
- Client
- None
01What it solves
Teams want AI agents to act on real systems without letting the model decide what it is allowed to do.
- Who it is for
- Engineering teams running AI agents on their own infrastructure.
02What was built
- Scoped capabilities that can only narrow
- Deterministic, deny-by-default policy engine
- Human approvals bound to one exact call
- Hash-chained, tamper-evident audit log
- Secret broker: secrets never reach the model
- Path jail, SSRF checks, no shell
- Anthropic, OpenAI-compatible and mock providers
svtCLI: doctor, policy check, approvals, audit
03In use

04Built with
- Python
- Pydantic
- Typer
- Rich
- httpx
- SQLite
- OpenTelemetry
- MCP
- pytest
- mypy
05Where it stands
In development
Version 0.1 (alpha): the foundation is complete and tested; not yet hardened for multi-tenant or internet-facing use.
- Started
- 1 October 2026
- Last updated
- 2 October 2026
06Code and links
- Code
- Private repository. The code is not public.
- See it
- No public deployment. The recorded film and screens on this page show it running.
Want AI in a workflow, with people keeping the decisions?
Tell us what you need and by when. We read every enquiry and reply in writing, with questions, a next step, or a plain no if we’re the wrong studio.



