Yash S | Founder | Building SecureVector, open source AI firewall for OpenClaw & Agents. Blocks prompt injection, controls tool permissions & caps API costs.
@ThePeterMick yep all to-do items are in my github project , easy to manage and claude code can directly refernce it from github project , this helps a lot.
OpenAI Codex can run shell commands, read environment variables, and call MCP servers.
It logs the commands and their output — but as plain text it can't enforce or tamper-proof anything.
SecureVector v4.4.0 adds that: SecureVector-Guard plugin runs inside Codex, checking (allow/deny) and logging every tool call to a tamper-evident audit chain, on-device.
Free + open source. #AIagents #infosec
I catalogued every MCP server on npm, PyPI, GitHub, and the official registry.
3,821 of them.
Most teams have no idea which ones their AI agents are talking to right now.
Built plugins for Claude Code and OpenClaw that enforce policy on every MCP tool call and catch prompt injections at the door.
Part 2 below. Part 1 in the reply.
https://t.co/zIUqluRm0B
#ClaudeCode #MCP #AISecurity
@natalie_avfieb Launched v4.3.0 with feature you requested with Bidirectional monitoring, please check my post below. thanks! Let me know if this is helpful https://t.co/HdFQGVRjsx
So many people building on OpenClaw with no security layer 👀
Monitor what your agents are interacting with, get full visibility into what’s being sent to your LLMs, and catch threats before they become problems.
Try it free 👇
https://t.co/10gvhf7HXQ
Any issues or feedback — happy to hear it !
#OpenClaw #AIAgents #LLMSecurity #n8n #AIFirewall #opensource #buildinpublic”
@Timur_Yessenov Launched v4.3.0 with feature you requested with "bill of tools", check my post below. thanks! Let me know if this helps or you would like to see anything more specific. https://t.co/HdFQGVRjsx
Fantastic feedback thanks a lot! I will work on building this …. calling it the “MCP Bill-of-Tools Digest”. Per-device export in the local app: server, source, last used, secrets-touch in one view.
Filing as v4.3.0… will post here as soon as available.
Let me know if you also interested seeing device fleet wide report in cloud app.
@TTrimoreau Built SecureVector, Realtime security and tool auditing layer for agents/harnesses ( like claude code, openclaw) , did not get any feedback yet but have 9k downloads so tough to answer
SecureVector v4.2.1 - Claude Code plugin landed 🚀
→ SecureVector plugin for Claude Code — every MCP tool call in your statusline, audited locally.
→ MCP Catalogue + Cloud Policies — author policies to allow/deny/audit tools, push it across every enrolled developer machine (or multiple devices) on next sync. cloud - https://t.co/lb4xvplsh9
local app- https://t.co/10gvhf7a8i
#ClaudeCode #MCP #AISecurity #OpenClaw #SecureVectorAI