What I use

Primary IDE.

Agent loop with sandbox subprocess isolation enabled (CLAUDE_CODE_SUBPROCESS_ENV_SCRUB=1, PID-namespace isolation). proofloop (the rubric I ship) is verified against this version.

Long-running sessions, paired with Cursor for review-on-cursor and ship-on-tmux.

Strategy + planning + nuanced long-context.

Default workhorse for the Agentify orchestrator.

Routing + pre-classification.

OpenAI default.

Hard-reasoning cases.

Eval-only, not in production rotation. Used to triangulate when Anthropic + OpenAI disagree.

Active eval. Reasoning-toggleable; agent-runtime evals only, not yet in production routing fallback chain.

FastAPI, Pydantic v2, asyncio everywhere.

Frontend + Edge runtimes.

App Router, RSC, Edge + Node runtimes per route. Pinned ≥ 16.3.6, the security patch floor CI enforces.

All frontend + most scripts.

mnemo + critical infra paths, anywhere I need tight memory control + predictable latency.

Hosts this portfolio: Next.js SSR behind CloudFront.

Frontend + Edge functions for the agent-airlock demo surfaces.

Long-running agent workloads at Attri.ai.

Limited preview, Apr 28 2026. Wraps any Bedrock Codex / OpenAI GPT invocation with the agent-airlock sandbox + capability-lease envelope. Optional bedrock_invocation_arn field shipped in /api/quote/schema.json v0.2.

Code execution sandbox for unverified agent-written code. Zero production incidents to date.

Real-time multi-tenant collaboration in Agentify.

Primary store. Vector search lives next to the source of truth.

Okta Universal Directory now hosts non-human identities (NHIs) for agents alongside humans. Lifecycle, audit, revocation are first-class. Opt-in module in the consulting starter from today; existing engagements get a one-line patch that binds capability leases to Okta NHI tokens.

Cryptographic identity layer. /api/quote already accepts agent_identity.did + verifiable_credential per draft-2020-12 schema; verification on the v0.4 roadmap.

Time-bounded, scope-bounded lease per call with audit-emit channel. Anthropic Project Deal · Cloudflare Mesh · Cisco AWI converge on this shape.

Type-checker and contract layer for agent tool calls, deny-by-default and in-process: strict argument validation, ghost-argument stripping, optional PII masking, RBAC, and sandboxed execution.

SAST scanner for agentic systems: OWASP Agentic Top-10 + MCP Top-10 + 14-framework compliance reporting (SARIF).

7-dimension quality rubric for Claude Code; auto-hooks into the agent loop.

Red-team & assurance for physical AI: attacks vision-language-action robot policies in simulation, reports an attack-success rate with a 95% CI + benign control. The action-layer counterpart to agent-audit-kit's text-layer scanning.

Tracing + metrics across the multi-agent fleet.

Error capture, frontend + backend.

Product analytics + feature flags.

Trace surface for agent steps.

Primary daily driver.

12+ hour days demand it.

Edge compute. JetPack 6.3 (L4T r36.4.x).

Depth + IMU sensing. Sourced, integration in progress.

Differential-drive base, 5-DOF arm + parallel gripper. In fabrication.

Migrating from Humble; Nav2 stack on top.

TensorRT-optimized OWL-ViT for open-vocabulary detection. Pinned commit-sha checked into the build manifest.

NVIDIA April 2026 release. Used as architectural reference, pinned to a specific commit-sha in the build manifest.

Synthetic-data factory for affordance training.

Simulation suites Provael red-teams VLA policies inside. Deterministic, CPU-first for CI.

Hugging Face VLA policies Provael targets. SmolVLA is the primary measured checkpoint.

Provael's Python 3.12 env + build toolchain: reproducible lockfile, fast CI.