TTRL / Repositories
TTRL AGENT RUNTIME / INTERACTIVE WALKTHROUGH

Intent is a proposal.
Policy is the boundary.

Explore how an agent request moves through policy, human approval and an audit trail. Built around the runtime’s documented Rust trust core and Go MCP gateway architecture.

Current project: MosinShaikh505/TTRL-Agent-Runtime ↗ · Private repository; GitHub access required.

Browser demoNo connected runtime, tools, secrets or signed receipts.
CONNECTED POLICY / SEPARATE FROM THE WALKTHROUGH

Evaluate the real policy boundary.

This connection uses fixed scenarios through the Rust policy bridge when its isolated API is available. It does not execute tools, issue human-approval grants or create full runtime-ledger receipts.

Live policy service is not connected. The walkthrough below is a browser simulation.

No connected evaluation has run.
CONTROL PLANE / DEMO

Govern every next step.

SIMULATION
Policy authorityRust trust core
Protocol boundaryGo MCP gateway
Actual dispatches0 / disabled
01 / PROPOSE AN ACTION

A bounded request.

Fixed sample rules. No model, network request or executable workload runs.

02 / DECISIONREADY

A request awaits review.

Select a sample request, then evaluate its illustrative policy.

EXTERNAL EFFECTSNone — dispatch remains disabled
03 / EVIDENCE WALKTHROUGH

Follow the decision.

  1. Request
  2. Policy
  3. Decision
  4. Demo evidence

Ready. Select a sample request.

  1. No demo events yet.

This is an unsigned browser event list. The real runtime’s signed audit evidence must be generated and verified in an actual configured environment.

DOCUMENTED ARCHITECTURE

Authority stays
outside the model.

The repository describes a gateway that validates requests, a Rust core that evaluates deterministic TAPL policy and approvals, bounded tool execution, and signed audit evidence. This walkthrough explains those boundaries without impersonating a live deployment.

01

Validate intent

Go MCP gateway and action firewall at the tool-request boundary.

02

Evaluate authority

Deterministic policy, scoped capabilities and bound approval continuations.

03

Constrain execution

Governed STDIO and Wasm brokers, with environment-dependent containment.

04

Inspect evidence

Signed dispatch, outcome and failure evidence in the actual runtime.

ENGINEERING STATUS / CLAIM BOUNDARY

Built for scrutiny.

The authoritative repository is MosinShaikh505/TTRL-Agent-Runtime. Its README was reviewed on 7 November 2026 and records seven unclosed parent exit gates: independent receipt trust, combined containment and quota pressure, identity-provider acceptance, operational replication/failover, genuine hardware quotes, installer platforms and final audit/documentation.

Customer-environment sign-off remains open. This page does not claim a completed production release, independent security certification, live remote administration or real hardware attestation.

INTERACTIVE BUYER LAB

Explore the workflow.
Inspect the evidence.

Fictional CRM, one-shot approval, local investigation, browser Ed25519 verification, journal archival and security-control models. Clearly labelled simulation; no live gateway or customer administration.

Launch the interactive buyer demo →