Mindburn Labs
Founding Cohort
RFC-001 Published
Star
···

Governedorganizationalexecution.Authorize.Govern.Prove.Everyconsequentialactionreceipted.

HELM is the execution authority layer for organizations. Start with the open execution kernel, add shared control through HELM Platform, and prove every consequential action with receipts, ProofGraph, and EvidencePacks.

See How It WorksApply for Early Access
PIPELINE ACTIVE0 evaluated

Request flows through governance → cryptographic receipt issued

LIVE
HELM Releasesdk/go/v0.1.1Feb 19, 2026 · sig ✓Site BuildPASSe3b0c442 · 7/7 gatesLab Run🔬 Research Scout · Mar 19, 2026Latest DemoGate Evaluationreceipts · allow/deny
success: OPEN SOURCEHELM OSS v0.1.1
npx @mindburn/helm
Verifies the latest HELM OSS release offline. Prints SHA-256 + receipt.
GitHub Verification details
accent: CONFORMANCE0 checks
L1Structured
24/24
L2Receipted + Chained
17/18
L3Verified
10/12
conformance targets
// 03

Proof you can run

proof: VERIFY
Evidence Pack Verifier
Upload or paste an evidence pack — verify signatures, replay the receipt chain, confirm integrity offline.
accent: EvidencePackneutral: L3
Run danger: GATE
Gate Simulator
See who is allowed to execute — enforced by policy, constraints, and proofs.
accent: Receiptneutral: L1
Run warning: POLICY
Policy Playground
Author and test governance policies — see how allow/deny decisions change with different constraints.
accent: Receiptneutral: L2
Run info: GRAPH
ProofGraph Explorer
Browse immutable execution history — filter, query, and export receipt subgraphs.
accent: EvidencePackneutral: L2
Run success: L1–L3
Conformance Runner
Run L1/L2/L3 conformance tests in-browser and export reports with badge payloads.
accent: ConformanceBadgeneutral: L3
Run accent: ALL
View all demos
Browse the full catalogue of interactive proof instruments.
Explore
// 04

Governed Execution Pipeline

Every AI action is proposed, policy-checked, executed, and cryptographically receipted. Hover any stage to see its invariant and proof output.

InvariantAction is structured as a typed proposal — never raw execution
EmitsCallProposal { schema, params, caller_id }
Kernel boundary
Proof surface
External
// 05

Research Workstreams

Six workstreams drive verifiable AI execution forward. Each produces standards, code, and cryptographic evidence.

A2AAgent-to-Agent ProtocolsEVDEvidence & CryptographyKRNKernel OS ArchitecturesVERVerification Tooling
HELM Kernel
Core execution kernel, proposal-centric architecture, proof graph construction.
neutral: CTOneutral: Engineerneutral: Auditor
Evidence Systems
Evidence packs, hash-linked receipt chains, verification tooling.
neutral: Engineerneutral: Auditor
Conformance
L0–L3 conformance levels, test suites, CI enforcement gates.
neutral: CTOneutral: Auditor
A2A Protocols
Agent-to-agent communication standards, interoperability, transport.
neutral: CTOneutral: Engineer
TITAN Organism
Reference system for governed trading — TypeScript brain, Rust execution.
neutral: CTOneutral: Engineerneutral: Auditor
Verification Tooling
CLI verifiers, visual regression, security scanning, demo contracts.
neutral: Engineerneutral: Auditor
// 06

Releases & research

Feb 2026success: Release
HELM OSS v0.1 — Public Release
The full open-source execution kernel. OpenAI-compatible proxy, ProofGraph, fail-closed PEP.
▸⊢ View Evidence / Verify
hashab494e…896e
verifynpx @mindburn/helm --id helm
Feb 2026accent: Research
ProofGraph Specification (RFC-001)
PrevHash chaining + LamportClock monotonicity for tamper-evident execution trails.
▸⊢ View Evidence / Verify
hash851587…be3d
verifynpx @mindburn/helm --id proofgraph
Feb 2026warning: Engineering
Bounded Compute — WASI/Wazero (RFC-004)
Sandboxed execution with no network, no filesystem, and compute limits.
▸⊢ View Evidence / Verify
hash0ea5f2…4987
verifynpx @mindburn/helm --id bounded
All releases
// 07

References & Proofs

1. Fail-closed by default ✓ Verified

2. Zero-dependency setup ✓ Verified

3. Offline-verifiable receipts ✓ Verified

4. 5–10 ms overhead per tool call ✓ Verified