SaaSHub helps you find the best software and product alternatives Learn more →
Top 23 TypeScript agent-security Projects
-
emilia-protocol
Authority control plane for autonomous work. EMILIA Gate enforces finite customer-owned mandates at protected executor boundaries; the open protocol keeps evidence verifiable.
Project mention: Two Independent CCS Implementations Achieve 14/14 Interoperability | dev.to | 2026-08-21EMILIA commit: https://github.com/emiliaprotocol/emilia-protocol/commit/5fb2eae9b20d17685e1ecbe061eb4ced03dbe64e
-
SaaSHub
SaaSHub - Software Alternatives and Reviews. SaaSHub helps you find the best software and product alternatives
-
agent-threat-rules
Open detection-rule standard for AI agent security threats — like Sigma, but for AI agents. Executable, testable rules for prompt injection, tool poisoning, context exfiltration and MCP attacks. Merged into open-source projects at Microsoft, Cisco, Gen Digital, MISP and FINOS. MIT-licensed.
Project mention: I Scanned 2,386 MCP Packages on npm. 402 Were Critical. Here's What I Found. | dev.to | 2026-03-22Everything is open source (MIT): ATR rules + PanGuard scanner
-
devbox
Open source autonomous engineering harness in your cloud. One sandbox per task, no real secrets inside. (by rudderlabs)
Project mention: Show HN: DevBox – Open-source engineering harness from RudderStack | news.ycombinator.com | 2026-11-03 -
Stroq
Find known secrets in your AI coding agent sessions (Claude Code, Codex, Cursor), then guard the next one with local hooks that trace risky commands to the content that dictated them. Open source, no telemetry.
Project mention: 5 numbers worth pulling from your own coding-agent sessions in 2026 | dev.to | 2026-11-09Stroq is a small open-source tool whose documentation describes local analysis of coding-agent sessions, text-level command provenance and known-secret matching. The project says it has no telemetry. The descriptions are documented capabilities, not the results of an independent security audit here.
-
mcp-observatory
Free, MIT-licensed MCP security testing from ClearFrame. Scan, audit, compare, and review locally; continue in the ClearFrame CLI or Mac app.
Project mention: MCP Observatory: Scan, Test, and Secure MCP Servers Before Agents Depend on Them | dev.to | 2026-07-12MCP Observatory helps teams test, secure, and monitor MCP servers before agents depend on them.
-
lelu
Open source authorization engine for AI agents. Confidence-aware gating · Human-in-the-loop review · Policy-as-code · Full audit trail (by Lelu-ai)
Project mention: Show HN: Lelu – gate OpenAI agent actions on confidence and prompt injection | news.ycombinator.com | 2026-06-24 -
agentshield-benchmark
Open benchmark for AI agent security tools — prompt injection, data exfiltration, tool abuse, provenance
Project mention: GentShield – Open benchmark of 6 AI agent security tools (537 test cases) | news.ycombinator.com | 2026-02-15 -
agent-shield
Multi-engine security scanner for AI agents, MCP servers & plugins — 13 engines, one report.
We built Agent Shield, a security scanner for AI agent tools, and used it to audit 17 of the most popular MCP servers — including official ones from Anthropic, AWS, Cloudflare, Docker, Brave, and Azure.
-
agent-chaperone
Screens an AI agent's tool calls before they run and tool results before the agent reads them. An MCP proxy plus a hooks adapter for a client's built-in tools.
Project mention: Show HN: Agent Chaperone – Screen AI agent tool calls and results with Jev | news.ycombinator.com | 2026-09-21 -
melra
Open-source local MCP runtime that checks AI-agent tool calls, gates risky actions on approval, verifies results, and avoids blind retries after timeouts.
Project mention: Show HN: Melra – pause ambiguous MCP writes instead of blindly retrying | news.ycombinator.com | 2026-10-20 -
abtars
abTARS is a personal agent built around Pi harness: persistent memory with abmind, messaging connectivity, skills, scheduled task, self healing and multi agent collaboration
Project mention: AbTARS – Self-hosted AI agent with persistent memory and 5-layer self-healing | news.ycombinator.com | 2026-06-07 -
jevc
Compile agent policy prose into deterministic verdict programs: narrow evidence questions for the model, the verdict computed in code. Install: npm i -g jev-compiler
Project mention: Jev compiler – Turn the rules your agent keeps ignoring into gates you can test | news.ycombinator.com | 2026-09-20 -
-
Project mention: Show HN: TrustVector – Trust evaluations for AI models, agents, & MCP | news.ycombinator.com | 2026-02-13
-
toolpoison
Security scanner for MCP servers. Detects tool poisoning, prompt injection hidden in tool descriptions, leaked API keys, and supply-chain risk across Claude, Cursor, VS Code, Windsurf and Zed configs. One command, no API key.
This document is maintained alongside toolpoison, a scanner that detects most of what is described here. It is written to be useful on its own.
-
mguard
Memory defense for AI agents — stops MINJA, AgentPoison, and MemoryGraft attacks. Zero dependencies.
Project mention: Show HN: Mguard – First defense against MINJA memory poisoning attacks | news.ycombinator.com | 2026-03-08 -
recued
A self-hosted gateway between AI agents and your email, CRM and other accounts. Agents get only the tools you grant, never the keys behind them, and by default anything that writes, sends or deletes waits for your yes.
Project mention: Ask HN: What are your most ambitious projects in 2026? | news.ycombinator.com | 2026-10-24I started planning for this project in February, began coding in April, first release in July, and been active daily till today.
At first I was only trying to make a tool that makes my day easier, I want AI to follow the exact procedure to perform my requests, soon I realize almost 80% of the "expected procedure" is deterministic without even needing AI, in most cases AI is only needed in judge/review/summarize/orchestrate actions, all the rest of the dirty work can be done without AI and without paying a dime.
As a result I created https://github.com/recued/recued, it stores all my data in my own home, executes predefined recipes whenever I ask for, fast, and cheaply.
Is it a success? I don't know, but I am still actively improving it on daily basis.
I guess neither success nor failure fits my description, I found something that I am passionated about, and it is also making my day easier.
From the sentimental perspective, I call it my success.
-
Synapsor-Runner
Let AI agents query and update Postgres/MySQL without raw SQL, unrestricted schema access, or database credentials in the model; writes stay reviewed.
Project mention: Let AI agents use production data without handing them your database | dev.to | 2026-07-16github.com/Synapsor/Synapsor-Runner
-
mcpcustoms
MCPCustoms is the customs checkpoint for MCP tool calls: a zero-dependency TypeScript runtime semantic firewall that inspects every Massive Intelligence (IM) agent tool call before execution. Detects tool poisoning, command/SQL/prompt injection, path traversal, and capability overreach; verifies signed Ed25519 publishers; fail-closed by default.
Project mention: These tools provide the engineering substrate required to meet the rigorous safety and economic constraints of production environments. | dev.to | 2026-06-28 -
diplomat-agent-ts
What can your TypeScript AI agent do to the real world? Scan your code. See which tool calls have zero checks
Project mention: Show HN: Scan your AI agent's code for tool calls with no checks | news.ycombinator.com | 2026-05-28 -
pod
pod — least-privilege compiler for AI agents. Compile a least-privilege policy from what your agent actually did, then prove what happened. (by suhui-organization)
Project mention: I scanned 14 MCP servers on my laptop, then compiled a least-privilege policy from what my agent actually did | dev.to | 2026-09-16Repo: https://gitee.com/suhuisoftwares/pod (mirror: https://github.com/suhui-organization/pod)
-
safelabs-guard
Runtime guardrail that stops multimodal agents from being fooled — Muse Spark + SAM 3.1 + Muse Voice verify perception against a deterministic ledger before action.
GitHub repository
-
stopline
A decision gate for browser agents: Jev or Laya classifies each proposed action, a policy in code decides whether it runs.
Project mention: Hacktoberfest 2026 Stopped Counting PRs. How I'd Spend the Month Instead | dev.to | 2026-10-02I try to hold my own small projects to that. Metron, an Apache 2.0 terminal coding agent that talks to a local model through Ollama, ships checksums, SBOMs and build-provenance attestations with each release, so you can verify the archive came from the repository's own build. It is not production certified, and the README says so. If you want something to read during the month, the Metron repository is open, and so is Stopline.
TypeScript agent-security discussion
TypeScript agent-security related posts
-
5 numbers worth pulling from your own coding-agent sessions in 2026
-
Show HN: Check your Claude Code, Codex and Cursor sessions for secrets
-
Show HN: Which of your secrets did your coding agent send?
-
285 Ways to Attack an AI Agent — A Security Taxonomy
-
Show HN: A 3-line wrapper that enforces deterministic security for AI agents
-
Meta hires duo behind Moltbook
-
Sandboxes won't save you from OpenClaw
-
A note from our sponsor - SaaSHub
www.saashub.com | 16 Nov 2026
Index
What are some of the best open-source agent-security projects in TypeScript? This list will help you:
| # | Project | Stars |
|---|---|---|
| 1 | emilia-protocol | 550 |
| 2 | agent-threat-rules | 429 |
| 3 | devbox | 330 |
| 4 | Stroq | 247 |
| 5 | mcp-observatory | 120 |
| 6 | lelu | 75 |
| 7 | agentshield-benchmark | 29 |
| 8 | agent-shield | 17 |
| 9 | agent-chaperone | 14 |
| 10 | melra | 13 |
| 11 | abtars | 10 |
| 12 | jevc | 9 |
| 13 | attestor | 8 |
| 14 | TrustVector | 5 |
| 15 | toolpoison | 4 |
| 16 | mguard | 4 |
| 17 | recued | 3 |
| 18 | Synapsor-Runner | 3 |
| 19 | mcpcustoms | 2 |
| 20 | diplomat-agent-ts | 1 |
| 21 | pod | 1 |
| 22 | safelabs-guard | 1 |
| 23 | stopline | 0 |