# paiml-mcp-agent-toolkit

> Pragmatic AI Labs MCP Agent Toolkit - An MCP Server designed to make code with agents more deterministic

- **URL**: https://www.freshcrate.ai/projects/paiml-mcp-agent-toolkit
- **Author**: paiml
- **Category**: MCP Servers
- **Latest version**: `v3.17.0` (2026-05-06)
- **License**: MIT
- **Source**: https://github.com/paiml/paiml-mcp-agent-toolkit
- **Homepage**: https://paiml.github.io/pmat-book/
- **Language**: Rust
- **GitHub**: 149 stars, 25 forks
- **Registry**: github
- **Tags**: `agentic`, `c`, `deno`, `kotlin`, `mcp`, `mcp-server`, `paiml`, `paiml-active-tool`, `rust`

## Description

Pragmatic AI Labs MCP Agent Toolkit - An MCP Server designed to make code with agents more deterministic

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v3.17.0` | 2026-05-06 | High | ## Highlights  Two ship-blocking bugs landed via #564 and #565, both discovered while building a Coursera RAG-from-Zero lab.  ### Fixed - **`scripts/install.sh` URL pattern, tarball layout, and Linux platform default** (#561 → #564). Three sub-bugs in the documented one-liner installer:   1. URL was constructed as `paiml-mcp-agent-toolkit-${PLATFORM}.tar.gz` but actual release assets are named `pmat-v${VERSION}-${PLATFORM}.tar.gz` — every install since the v3.0 rename returned 404.   2. The rele |
| `v3.16.0` | 2026-04-27 | High | ## Highlights  - **4 bug fixes**: `pmat comply` default-yaml severity defect (#527), rustls-webpki RUSTSEC-2026-0104 bump (#476), incremental analysis flake (#449), entropy regex word-end bug (#396) - **80% line coverage milestone** across the workspace - **354 new tests** added since v3.14.0 - **MCP**: forward `ToolInfo` metadata for 4 `pmat_*` handlers (#389) - **Verification**: Kani proof harnesses for pure functions (#299) - **Brew**: rename `homebrew/` → `Formula/` for tap convention (#388) |
| `v3.14.0` | 2026-04-15 | High | ## What's Changed  Migrate all sovereign stack dependencies from old crate names to the unified `aprender-*` 0.30 monorepo on crates.io.  ### Dependency renames  \| Old name \| New name \| Version \| \|----------\|----------\|---------\| \| `trueno` \| `aprender-compute` \| 0.30 \| \| `trueno-db` \| `aprender-db` \| 0.30 \| \| `trueno-graph` \| `aprender-graph` \| 0.30 \| \| `trueno-rag` \| `aprender-rag` \| 0.30 \| \| `trueno-viz` \| `aprender-viz` \| 0.30 \| \| `trueno-zram-core` \| `aprender-zram-core` \| 0.30 \| \| `present |
| `v3.10.0` | 2026-03-31 | Medium | ## What's New  ### `pmat explain` Command Look up any check, metric, or grade to understand what it checks, why it matters, and how to fix failures.  ```bash pmat explain CB-1210     # Explain a specific check pmat explain CB-12       # Prefix match: all CB-12xx checks pmat explain precondition # Fuzzy search in names/descriptions pmat explain             # List all available checks ```  19 check explanations across comply (CB-xxx), provable contracts (PV-xx), and TDG grades.  ### Provable Contr |
| `v3.8.0` | 2026-03-23 | Medium | ## Provable Contracts Integration  Escape-proof contract enforcement pipeline across the entire sovereign AI stack. YAML contracts → build.rs env vars → `#[contract]` proc macro → `debug_assert!()` injection. Zero runtime cost in release builds.  ### New Compliance Checks  \| Check \| What it verifies \| \|-------\|-----------------\| \| **CB-1203** \| Contract equations with preconditions have `#[contract]` macros \| \| **CB-1204** \| `build.rs` emits PRE/POST env vars from YAML contracts \| \| **CB-1205** |
| `v3.7.0` | 2026-03-09 | Low | ## Highlights  - **RPS v3.0**: New Reproducibility scorer wrapping Popper categories B-F (15 pts), falsifiability gateway (Jidoka), 11 categories / 289 max points - **PMAT-510**: Six scoring system improvements — Five Whys v2 evidence weights, Muda file mapping, EvoScore CB-142, `--rank-by priority`, `test --record` - **Performance**: -2.57 GB peak memory (deep context), -59% index build allocations, -44 MB graph clone elimination - **New commands**: `ci-local`, `bottleneck`, `test-stability`, ` |
| `v3.3.0` | 2026-02-15 | Low | ## What's New  ### Rich `pmat extract --list` output (#217)  `pmat extract --list` now returns a rich JSON object instead of a flat array, providing everything needed to split large files into compilable parts:  **New top-level fields:** - `file` — file path - `language` — detected language (`rust`, `typescript`, `python`, `c`, `cpp`, `go`, `lua`) - `imports` — all top-level import/use statements (full text) - `cfg_test_line` — line where `#[cfg(test)]` appears (Rust only)  **New per-item field: |
| `v3.1.0` | 2026-02-14 | Low | ## What's New  ### Lua Compliance Suite (CB-608 to CB-619) - **12 new Lua checks**: coroutine misuse, require cycles, global protection, weak tables, test framework detection, OpenResty patterns, FFI safety, OOP patterns, type annotations - Lua dead code detection with module export awareness - Lua mutation testing adapter with busted integration - Tree-sitter TDG analyzer for Lua with full 7-component scoring  ### SQL, Scala, YAML, Markdown, MLOps Compliance - **CB-700**: SQL best practices (N+ |
| `v3.0.7` | 2026-02-12 | Low | ## Fix #173: check_provability hardcoded score + config ignored  ### Changes - **Real provability analysis**: `check_provability()` now calls `calculate_provability_score()` using the `LightweightProvabilityAnalyzer` instead of returning a hardcoded `0.65` - **Config threshold**: Added `provability_min = 0.60` to `.pmat-metrics.toml` so the threshold is read from config instead of always falling back to the 0.70 default  ### Files Modified - `src/cli/analysis_utilities/quality_checks_part2.rs` — |
| `v3.0.6` | 2026-02-12 | Low | ## What's New  ### CB-600 Lua Best Practices Detection (PMAT-487) 8 new compliance checks for Lua codebases, grounded in LuaTaint, FLuaScan, Luau type system, and luacheck research:  - **CB-600**: Implicit globals — assignment without `local` keyword - **CB-601**: Nil-unsafe access — chained calls on function returns, deep field chains - **CB-602**: pcall error handling — uncaptured/unchecked pcall/xpcall - **CB-603**: Deprecated/dangerous API — `os.execute()`, `loadstring()`, `setfenv()` - **CB |

## Citation

- HTML: https://www.freshcrate.ai/projects/paiml-mcp-agent-toolkit
- Markdown: https://www.freshcrate.ai/projects/paiml-mcp-agent-toolkit.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/paiml-mcp-agent-toolkit/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._