@codapult/guard is local-first, model-agnostic architecture guardrail software for AI coding agents. Use it to detect architectural drift, enforce approved software architecture rules, and check code changes in JavaScript and TypeScript projects.

Fastest enterprise AI gateway (50x faster than LiteLLM) with adaptive load balancer, cluster mode, guardrails, 1000+ models support & <100 µs overhead at 5k RPS.

Exposes an assess_repo tool that scans a repository, local folder, or npm/PyPI package and returns supply-chain risk findings with file, line, and code evidence — install-time scripts, dangerous patterns, secrets, AI-agent manipulation, impersonation and published-artifact mismatches — while never executing the target's code. Findings carry severities from info to critical so results can gate installs and CI jobs.

Exposes agent-facing tools that let a personal AI agent submit each proposed action for an allow, require_approval, or deny verdict against OWASP PAA10 policy rules, plus tools to record provenance-tagged memories, log refusals, read status, and trigger a self-stop. Its decisions are advisory unless paired with the project's gating proxy or hook, which enforce them.

Enables agents and applications to detect, inspect, and neutralize prompt injection, jailbreak attempts, and hidden instruction escapes in untrusted text before it is passed into prompts.

Find and redact leaked secrets, wallet keys and PII before an AI agent sends text.

Check an email before your AI agent sends it: allow, review or block with reasons.

🛡️ Security audit CLI for Model Context Protocol (MCP) servers — scan AI agent configs for tool poisoning, rug pulls, hardcoded secrets, command injection & supply-chain risks. Pure Python, SARIF + CI ready.

High-frequency in-browser behavior tree execution engine for autonomous AI agents — 60Hz tick loops, reactive triggers, telemetry recordings, and safety guardrails.

A free, read-only server that lets users diagnose production AI-agent risks: it flags over-privileged API keys from permission descriptions, scans code snippets for the signals that prevent duplicate execution, and explains guard patterns such as idempotency keys, approval gates, outboxes, and circuit breakers. It makes no network calls or writes, and rejects inputs that look like real secrets.

Enables an AI agent to check live GPU prices, get a cost quote, rent a per-minute GPU instance, upload code and data, run jobs in the background, tail logs, download results and destroy the instance to stop billing. Includes guardrail limits on spend, hours and idle time, safe file transfer, and access to an OpenAI-compatible model API for inference without renting a GPU.

Guardian agent for AI coding: four frontier models review risky diffs and commits before they ship.

Guardrails for LLM output: pass / fail / review verdicts and one recommended action. Hosted or npm.

Enables any MCP-connected agent to drive a browser by reading each page as an indexed action-space table and performing one guarded operation on a numbered target, refusing and refreshing when the page changes instead of retrying selectors.

Gives AI agents a real Chromium browser via Playwright to QA-test web apps, with pass/fail assertions, network mocking, recorded traces and reports, and safety guards that block live payment traffic. Test outcomes come back as explained verdicts suitable for review or email delivery.

Embeds a runtime MCP inspector in a NestJS app so coding agents can inspect the module and provider graph, routes and their guards/interceptors/pipes/filters, logs, request traces, event listeners, and scheduled jobs through an HTTP endpoint.

Enterprise AI Platform with guardrails, MCP registry, gateway & orchestrator

Enables agents and terminal users to inspect ClickHouse over HTTP by listing databases and tables and running guarded read-only SELECT queries without modifying data.

Gives an existing AI agent local stdio tools to list, read, create, and edit .kinoscript.json audio-visual scripts in a single folder on your machine. Writes are validated, atomic, confined to that folder, and optionally guarded by an updated-at check so concurrent edits aren't silently overwritten.

Prove that what already worked still works after an agent changed the code. Picture checks, guards for fixed bugs, a pre-release walkthrough, and known-good markers — as a CLI and as an MCP server.

Enables AI agents to drive a real Chromium browser by receiving compact, token-efficient page maps and only the changes after each action. It guards against hidden prompt injections, blocks payments, sends and deletions until the user approves, and lets agents learn site-specific notes over time.

Exposes the full Plexim PLECS Standalone JSON-RPC interface as MCP tools, letting agents auto-start and connect to PLECS to load and edit models, run simulations and analyses, manipulate scopes and webserver data, and perform code generation or FMU export with structured, size-aware results. Includes write guardrails, concurrency handling for long simulations, and batch/raw passthrough escape hatches.

Exposes a governed dbt MetricFlow semantic layer to AI agents through three tools — list metrics, list dimensions and query metrics — so questions about federal contract spend are answered from defined metrics rather than raw tables. Guardrails reject undefined metrics, raw columns, SQL in filters and bulk exports, and no tool can execute SQL.

Enables learners to browse, read, and be quizzed on saved lessons inside Claude Desktop and Claude Code, and enables five Claude agents to build fact-checked K–College lessons through guarded, pinned MCP tools.