The first IDE where every agent has a TrustScore from minute one. AgentFoundry ships TrustModel's AGP Edge sidecar as a default — governance is no longer a vendor add-on, it's a property of the development surface.
AgentFoundry and TrustModel ship complementary halves of the same problem: developers building agents need a dev surface; CISOs deploying agents need a governance surface. We share a founder and a roadmap — this is a 1+1=3 motion, not a logo swap.
AgentFoundry owns the agent-build IDE — Claude Agent SDK, Cursor SDK, Microsoft Agent Framework, Google ADK, replay, fork, ship. TrustModel owns the AI Assurance Control Plane — Observe, Protect, Assure across 10 trust dimensions. Neither stack overlaps; both stacks need the other to close the loop.
Both companies are founded and led by Karl Mehta. The TrustModel AGP Edge contract, the AgentFoundry sidecar runner, and the IPC bridge were designed by the same person against the same spec — there is no integration tax, no quarterly partner negotiation, and no version drift.
Every AgentFoundry user lands a governed-by-default agent — which becomes a TrustModel telemetry source on first run. Every TrustModel customer gets a recommended IDE that natively respects their policies — which becomes AgentFoundry distribution. Distribution flywheel: both products grow off the same install.
The AgentFoundry ↔ TrustModel integration is live code on the QA channel today. Four landed PRs, one verified API contract, and a Fleet view that already speaks to TrustModel's control plane.
commit c6c5cea
commit 72c2dd5
agp-edge launcher CLI so the sidecar can be started outside of AgentFoundry — same image, same bootstrap flow. Enables headless agents, CI environments, and TrustModel customers who want to dogfood the runner without installing the full IDE. commit 592d00b
/install response schema verified live on QA
End-to-end smoke against api-trustmodel.pdxqa.com — bootstrap-token issued, sidecar enrolled, fleet record created with a live TrustScore reference. AgentFoundry's contract types are now generated against the QA-verified schema, not a draft. commit e523780
TrustModel's published Control Plane stack lands inside AgentFoundry with no architectural reshuffling. Each layer maps to a concrete AgentFoundry surface we already ship.
AgentFoundry emits agent-run telemetry — model invocations, tool calls, replay traces — straight into TrustModel's TelemetryTraceStrategy. The same trace that powers AgentFoundry's replay UI feeds the TrustModel Observe layer.
The AGP Edge sidecar runs alongside the AgentFoundry agent process. Bootstrap-token enrollment binds the sidecar to the customer's TrustModel tenant. Tool calls and outbound model invocations flow through the local Edge daemon for policy enforcement.
Every agent row in AgentFoundry's Agent Fleet view will carry a live TrustScore from TrustModel — across the 10 dimensions (safety, fairness, accuracy, privacy, transparency, robustness, accountability, explainability, compliance, reliability). One number a developer can defend.
The same enrollment event creates a AgentFoundry user and a TrustModel-managed agent. Both teams sell into the same install — no attribution war, no cannibalization.
When a developer starts an agent in AgentFoundry, the AGP Edge sidecar boots by default. The agent is enrolled into the customer's TrustModel tenant with zero extra configuration. Day-0 trust governance, no sales motion required.
CISOs deploying TrustModel ask "which IDE respects my policy?" The TrustModel answer becomes AgentFoundry — the only IDE that natively boots AGP Edge and surfaces TrustScore in the build loop. AgentFoundry becomes the distribution channel for governed development.
Tracked under TRUS-1042 (AGP Studio Integration) and TRUS-1044 / TRUS-987 in the TrustModel Jira. Engineering is staffed; UX is iterating on the Fleet view this sprint.
/install contract.
This page is for the TrustModel engineering and GTM team. The product integration is live on QA; the next conversation is co-marketing, joint sales motions, and the launch surface for "Governed agent development, built-in." Let's pick a thirty-minute window.
Email Karl directly