Complete 7-service sovereign stack (PostgreSQL 18, OpenBao, Keycloak, Gitea, SeaweedFS, OPA, Control Plane) idle memory.
platform/deploy/appliance/sizing-budget.toml
Other coding agents delete failing tests, collide across branches, and burn 3.5x more tokens on raw grep. Mindaeon enforces host-side test integrity (100% interception), runs 3.77 ms AST mutation testing, coordinates cross-repo contract sagas, and ships across CLI, Desktop, VS Code, Browser Extension, and Office 365.
$ mindcode login https://mindaeon.ai
✓ Authenticated via OIDC + Device Attestation (Ed25519)
✓ Hybrid Code Intelligence index ready (Tree-sitter + SCIP + pgvector)
$ mindcode session run --verify-mutations "Harden tenant billing webhook handler"
[00:01.2] Retrieved 6 symbols via RRF (71.2% fewer prompt tokens than grep)
[00:04.8] Applied patch to platform/apps/control-plane/src/commerce/webhooks.ts
[00:05.1] Host Test Integrity Guard: 0 skipped tests, 0 deleted assertions (PASS)
[00:05.4] Changed-Surface AST Mutation Gate: 14/14 mutants killed in 3.77 ms (PASS)
[00:05.5] Signed SLSA v1.0 in-toto attestation -> commit 8f41c09
Mindaeon Desktop · Worktree Collision Radar (Active Worktrees: 4)
─────────────────────────────────────────────────────────────────────────
Worktree A (feat/oauth-pkce) ──┐
├─► Shadow Merge Dry-Run: CLEAN (0.01 ms)
Worktree B (fix/rls-tenant-card) ──┘
Worktree C (refactor/billing) ──┐
├─► ⚠ AST Symbol Overlap Detected:
Worktree D (feat/stripe-tax) ──┘ src/commerce/invoices.ts#settleInvoice
→ Auto-serialized before merge conflict!
Result: 76.0% reduction in parallel agent merge conflicts (25 -> 6).
Cross-Repository Contract Radar & Atomic Rollback Saga
─────────────────────────────────────────────────────────────────────────
1. Producer Repo:billing-service (OpenAPI /v1/invoices schema change)
2. Consumer Repo:console-ui (TypeScript SDK client)
3. Consumer Repo:mobile-app (GraphQL federated resolver)
[Contract Radar] Detected breaking field removal on Invoice.taxExempt (100% recall)
[Campaign Saga] Generated coordinated PRs in topological order (Producer -> Consumers)
[CI Verification] Consumer #2 failed integration check -> Atomic Saga Rollback triggered
[Status] All 3 repositories restored to consistent state in 1.4s.
$ sudo mindaeon-appliance install --bundle ./mindaeon-appliance-0.1.0.tar.zst
✓ Verified offline image digests & CycloneDX SBOM (0 CVEs across 4 images)
✓ Unsealed OpenBao KMS transit engine (Ed25519 + AES-256-GCM)
✓ Migrated PostgreSQL 18 + pgvector with FORCE ROW LEVEL SECURITY
Service State Memory Budget
──────────────────────────────────────────────────
postgres healthy 264 MiB 512 MiB
keycloak healthy 418 MiB 640 MiB
control-plane healthy 142 MiB 384 MiB
openbao + opa healthy 89 MiB 256 MiB
gitea + seaweed healthy 178 MiB 456 MiB
──────────────────────────────────────────────────
TOTAL IDLE 1,091 MiB 2,248 MiB (48% of budget)
Every benchmark below is continuously executed by our in-repository evaluation suites
(bun run evals) and enforced in CI.
Complete 7-service sovereign stack (PostgreSQL 18, OpenBao, Keycloak, Gitea, SeaweedFS, OPA, Control Plane) idle memory.
platform/deploy/appliance/sizing-budget.toml
Blocks test file deletion, .skip/.only injection, assertion removal, and snapshot tampering outside the agent sandbox.
platform/evals/suites/verification-gate.eval.ts
Deterministic changed-surface AST mutations verify that new tests genuinely fail when production logic is broken.
platform/evals/suites/verification-gate.eval.ts
Hybrid Tree-sitter AST + SCIP/LSP call graph + pgvector semantic embeddings fused with Reciprocal Rank Fusion.
platform/evals/suites/code-intelligence.eval.ts
Continuous Shadow Merge dry-runs across parallel git worktrees detect overlapping AST symbols in p95 = 0.01 ms.
platform/evals/suites/worktree-radar.eval.ts
Shadow model replay and deterministic prompt prefix ordering maximize cache hit rates without quality regression.
platform/evals/suites/finops-shadow.eval.ts
Competitors force a false choice between cloud-locked autonomous agents that can tamper with tests and limited self-hosted chat plugins. Mindaeon delivers full multi-worktree autonomy inside a cryptographically governed, air-gap-ready platform.
| Capability / Guarantee | Mindaeon v0.1.0 | Claude Code | Cursor 2.0 | Devin 2.2 | Factory | GitHub Copilot | GitLab Duo |
|---|---|---|---|---|---|---|---|
|
True Air-Gapped Single-Host Appliance
Sovereignty
|
✓ VerifiedVerified: 1,091 MiB idle RAM (PostgreSQL 18, OpenBao, Keycloak, Gitea, SeaweedFS, OPA) | ✕ NoCloud-only control plane; requires outbound Anthropic/cloud API connectivity | ✕ NoProprietary SaaS backend required for indexing, routing, and Background Agents | ✕ NoCloud-hosted Firecracker VMs orchestrated by Cognition SaaS | ✕ NoMulti-tenant SaaS orchestration plane | ✕ NoGitHub Enterprise Cloud required for Copilot Coding Agent | ◐ PartialSelf-hosted model gateway available, requires multi-node GitLab Ultimate |
|
Zero Outbound Telemetry by Default
Sovereignty
|
✓ Verified0 external calls; 100% local hash-chained audit logs & OpenBao transit encryption | ✕ NoTelemetry & usage metadata transmitted to vendor cloud | ✕ NoPrivacy mode still routes requests through Cursor cloud proxy | ✕ NoSession transcripts and trajectories stored in Cognition cloud | ✕ NoCloud session state and analytics retained by vendor | ✕ NoRequires continuous GitHub Cloud license & telemetry check-ins | ◐ PartialConfigurable on self-managed instances with air-gapped license |
|
Host-Enforced Test Integrity Guard
Verification
|
✓ Verified100% (7/7) interception of test deletion, .skip/.only, and assertion weakening outside sandbox | ✕ NoAgent shares write access to test files inside working tree | ✕ NoAgent can modify or skip failing tests unless manually reviewed | ✕ NoFull VM write access allows modifying test suites to pass CI | ◐ PartialPrompt-level guardrails without host-enforced AST diff verification | ✕ NoRelies on downstream PR review to catch weakened tests | ✕ NoNo AST-level test tampering detector |
|
Changed-Surface AST Mutation Testing
Verification
|
✓ VerifiedDeterministic AST mutants verified in p95 = 3.77 ms (100% weak-test rejection) | ✕ NoNot included | ✕ NoNot included | ✕ NoNot included | ✕ NoNot included | ✕ NoNot included | ✕ NoNot included |
|
Cross-Repo Contract Radar & Atomic Rollback Sagas
Multi-Repo
|
✓ Verified100% breaking-change recall across OpenAPI/GraphQL/Protobuf/TS with atomic rollback sagas | ✕ NoSingle-repository working directory scope | ◐ PartialMulti-root workspace indexing without topological PR saga rollback | ◐ PartialMulti-repo sessions without deterministic contract AST verification | ◐ PartialCross-repo Droids without cryptographic saga rollback guarantees | ✕ NoScoped to single repository per issue/PR workflow | ✕ NoScoped to single project merge request context |
|
Parallel Worktree Collision Radar & Shadow Merge
Orchestration
|
✓ VerifiedContinuous background dry-run merges cut parallel agent conflicts by 76% (25 -> 6) | ◐ PartialSupports git worktrees manually; no live cross-worktree shadow merge radar | ◐ PartialUp to 8 parallel worktrees; conflicts resolved only at merge time | ✕ NoIsolated cloud VMs; conflicts surface only after PR creation | ✕ NoCloud task branches without real-time AST symbol overlap detection | ✕ NoIsolated GitHub Actions branches without cross-branch radar | ✕ NoNo worktree orchestration |
|
Hybrid Code Intelligence (Tree-sitter + SCIP + pgvector RRF)
Intelligence
|
✓ Verified71.2% prompt token reduction and 94% symbol recall via Reciprocal Rank Fusion | ◐ PartialAgentic ripgrep/glob search (high token burn on large monorepos) | ✓ VerifiedCloud-hosted Merkle-tree embeddings + LSP context | ◐ PartialDeepWiki cloud indexing | ◐ PartialCloud repository graph indexing | ◐ PartialGitHub Blackbird code search + embeddings | ◐ PartialZoekt exact code search + basic RAG |
|
Deterministic Shadow FinOps & Prompt Cache Optimizer
FinOps
|
✓ Verified50.0% verified cost reduction via shadow replay & prefix cache stabilization | ◐ PartialAnthropic prompt caching supported; no multi-provider shadow replay | ✕ NoOpaque credit-based model routing pool | ✕ NoOpaque ACU consumption billing | ✕ NoStandard token/seat billing without shadow model evaluation | ✕ NoPremium request multiplier quotas | ✕ NoFixed add-on seat licensing |
|
SLSA v1.0 / in-toto Provenance & 9 Compliance Packs
Compliance
|
✓ VerifiedEd25519-signed session provenance + SOC2, ISO27001, HIPAA, FedRAMP, PCI, GDPR, DORA, NIS2, EU AI Act | ✕ NoNo built-in SLSA v1.0 per-commit agent provenance | ✕ NoSOC 2 Type II vendor cert only; no signed diff provenance | ✕ NoEnterprise audit logs without cryptographic hash chains | ◐ PartialEnterprise audit trails without in-toto attestations | ◐ PartialGitHub audit log events without SLSA agent attestations | ◐ PartialGitLab compliance center integration |
|
Unified 6-Surface Parity (CLI, Desktop, IDE, Browser, Web, Office 365)
Surfaces
|
✓ Verified289/289 parity across Mindcode CLI, Desktop, VS Code, Browser Extension, Console & Office Add-in | ◐ PartialCLI, Desktop, IDE, and Web (no native Office 365 Add-in or self-hosted control plane) | ◐ PartialIDE fork, CLI, and Web agent (no Browser MV3 extension or Office Add-in) | ◐ PartialWeb workspace, Slack, and IDE integration | ◐ PartialWeb console, CLI, and IDE bridge | ◐ PartialIDE, CLI, GitHub Web, Mobile | ◐ PartialIDE, GitLab Web UI, CLI |
Download the surface that fits your workflow, then sign up to connect all of them to your organisation's shared context store, model broker, and audit chain.
Multi-worktree command center with live Collision Radar, Shadow Merge dry-runs, integrated PTY terminal multiplexer, visual AST diff review, and interactive Mermaid codemaps.
Manifest V3 extension that bridges live browser DOM state, console errors, network waterfall timings, and GitHub/GitLab pull requests directly into your active Mindaeon session.
Inline diff review, workspace symbol graph navigation, MCP tool approvals, and background task handoff directly inside Visual Studio Code or any OpenVSX-compatible editor.
Zero-dependency self-contained binary compiled for Linux (x64/ARM64 glibc & musl), macOS (Apple Silicon & Intel), and Windows x64 with 289/289 parity.
Provision your organisation in seconds, configure model providers and budgets, author Rego policies, and generate cryptographically signed compliance evidence packs.
Draft architecture RFPs, security questionnaires, and financial models in Word, Excel, and Outlook grounded in your live repository contracts and compliance evidence.
Every artifact is verified against our Ed25519 release manifest
(/downloads/manifest.json + /downloads/manifest.sig). Downloading any
client automatically opens the Platform Sign-Up onboarding step so you can link your device.
| Package & Artifact ID | Surface | Version | Size | SHA-256 Digest | Download |
|---|---|---|---|---|---|
|
Mindaeon Desktop — Debian / Ubuntu (.deb, x64)
desktop-0.1.0-linux-x64-deb
|
desktop | v0.1.0 | 126.4 MB |
f9151084c32fd8c29438…
|
Download (mindaeon_0.1.0_amd64.deb) |
|
Mindaeon Desktop — Linux Portable Tarball (.tar.gz, x64)
desktop-0.1.0-linux-x64-tar
|
desktop | v0.1.0 | 160.9 MB |
53f38352babee13f2ab7…
|
Download (mindaeon-0.1.0-linux-x64.tar.gz) |
|
Mindaeon Browser Extension — Chrome / Edge / Brave (MV3 .zip)
browser-extension-0.4.0
|
browser-extension | v0.4.0 | 120.2 KB |
fb0aaf222573ce37007f…
|
Download (mindaeon-browser-extension-0.4.0.zip) |
|
Mindcode VS Code / Cursor / VSCodium Extension (.vsix)
vscode-0.1.0
|
vscode | v0.1.0 | 4.3 MB |
bd54ac9ebb43f56c8684…
|
Download (mindcode-0.1.0.vsix) |
|
Mindcode CLI Runtime — Linux x64 (glibc)
runtime-0.1.0-linux-x64-baseline
|
runtime | v0.1.0 | 89.3 MB |
0aa8f4f1d121fca0ce29…
|
Download (mindcode) |
|
Mindcode CLI Runtime — Linux ARM64 (glibc)
runtime-0.1.0-linux-arm64
|
runtime | v0.1.0 | 88.9 MB |
2a2f6056b752fb1248c0…
|
Download (mindcode) |
|
Mindcode CLI Runtime — Alpine Linux x64 (musl)
runtime-0.1.0-linux-x64-musl-baseline
|
runtime | v0.1.0 | 83.1 MB |
05354cf146faa8c122cd…
|
Download (mindcode) |
|
Mindcode CLI Runtime — Alpine Linux ARM64 (musl)
runtime-0.1.0-linux-arm64-musl
|
runtime | v0.1.0 | 81.6 MB |
7d90cc6121f3a7fbefe2…
|
Download (mindcode) |
|
Mindcode CLI Runtime — macOS Apple Silicon (M1–M4, arm64)
runtime-0.1.0-darwin-arm64
|
runtime | v0.1.0 | 70.3 MB |
cd3efd3d4c5984193855…
|
Download (mindcode) |
|
Mindcode CLI Runtime — macOS Intel (x64)
runtime-0.1.0-darwin-x64-baseline
|
runtime | v0.1.0 | 77.8 MB |
7981a06f4f7cb6fa8c14…
|
Download (mindcode) |
|
Mindcode CLI Runtime — Windows x64 (.exe)
runtime-0.1.0-windows-x64-baseline
|
runtime | v0.1.0 | 94.7 MB |
45aed65f9ac6b6363eef…
|
Download (mindcode.exe) |
release-dev-2026.Uwh1qAyuuYK9IawlL4fu9Y79tbpA_UVBBQwfgzQHhGbG3zM-v4BvRhuUN8iaTIxws6Hb6Pz_7A3xEQA_STr1BA
Audited continuously by 8 open-source security scanners across source code, dependencies, secrets, shell scripts, and all 4 container images (0 CVEs).
Verified with Syft CycloneDX SBOM generation and OSV-Scanner across control-plane,
runner, agent, and backoffice images.
Every tenant's secrets and session records are encrypted under isolated OpenBao transit keys
with PostgreSQL 18 FORCE ROW LEVEL SECURITY.
One-click cryptographic evidence bundles for SOC 2, ISO 27001, HIPAA, FedRAMP, PCI-DSS 4.0, GDPR, DORA, NIS2, and EU AI Act.
Every new organisation gets a 5-seat card-free trial with $5.00 of included model usage, full Desktop/Extension/CLI access, and instant tenant isolation.
5 seats included, $5.00 model usage credit, full Desktop app, Browser Extension, VS Code extension, and Mindcode CLI access. No credit card required.
Dedicated tenant subdomain, Parallel Worktree Collision Radar, Cross-Repo Contract Sagas, Shadow FinOps optimizer, and SAML/OIDC SSO.
Deploy inside classified, defence, financial, or sovereign VPC environments on a single 4 GB host or Kubernetes cluster with zero external network connectivity.