The Ecosystem
MIND is built on an open-core model. The language and compiler are fully open, while specialized runtimes power enterprise scale.
The official Rust implementation of the compiler. Includes the CLI, type checker, MIR/MLIR lowerings, and CPU executor.
- Built-in CLI & Type Checker
- MIR/MLIR Optimization Pipeline
- Native CPU Execution
The authoritative language specification, design documents, and RFCs. Changes to the language start here.
- Formal Language Specification
- RFC Design Documents
- Community Governance
Persistent, auditable memory for AI agents. Powers the Cognitive Kernel’s Memory Plane with versioned storage, contradiction detection, and cryptographic audit chains.
- Hybrid BM25 + Vector Retrieval
- MIND Scoring Kernels
- Cryptographic Audit Chain
- Drift Detection & Causal Graph
- Native MIC/MAP Wire Format
- 77 MCP Tools, Zero Runtime Deps
Deterministic architectural-governance scanner. Nine Q16.16 fixed-point metric kernels over a typed dependency graph — bit-identical across machines, cryptographically chained session deltas, rules expressed as MIND compile-time invariants.
- 9 Q16.16 Metric Kernels
- 5 Languages via tree-sitter + MIND reflection
- HMAC-SHA256 / Ed25519 Evidence Chain
- Bit-Identical Reproducibility
Production runtime with GPU acceleration, distributed training, compliance tooling, streaming execution, and safety-critical mode.
- CUDA GPU & Distributed Training
- Compliance & Audit Toolkit
- BCI / Safety-Critical Runtime
- Enterprise Support & SLAs
Graph authoring and compilation — compiled AI systems on top of the MIND stack.
- Typed
.flow.mindDSL — reuses the MIND type system - Compiles to native ELF via mindc → MLIR → mind-runtime
- Seven node kinds: native compute, LLM, tool, branch, memory, verify, plus agent delegation
- Deterministic trace; mind-mem as the memory plane
These components power the MIND Cognitive Kernel — a deterministic AI runtime with Control, Memory, and Verification planes.
Read the architecture docs