Launched this week
Kastra
Runtime authorization for Claude, Cursor, Codex and OpenClaw
817 followers
Runtime authorization for Claude, Cursor, Codex and OpenClaw
817 followers
Kastra is the runtime authorization layer for AI agents. It decides what agents can and cannot do before actions execute, enforcing policies with sub-1 ms latency across tools, prompts, inputs, and outputs. Use one control plane to govern agents and policies across Claude Code, Cursor, Codex, OpenClaw, the Anthropic SDK, the OpenAI SDK, and more. Prevent unauthorized tool use, prompt injection, and exposure of sensitive data before they become incidents. Trust the rules, not the agents.
Checking shell commands and tool calls before they run is the part that clicks for me. The local policy checks and scan of past agent activity make this feel practical, not just another monitoring dashboard.
Kastra
@adriancia Thanks! That was exactly our goal. We wanted to move from observing what agents did after the fact to deterministically deciding what they're allowed to do before they execute. The scan is there to make creating those policies much easier.
minimalist phone: reduce your screentime
So basically it prevents to disasters made by AI agents?
Kastra
@busmark_w_nika Exactly! No accidental file deletions, dropped database tables, emails sent to the wrong people, and so on. Kastra stops those actions before they execute, rather than reading about them in the logs afterward.
minimalist phone: reduce your screentime
@fernando_vintacourt That's cool, at least it will prevent some unsolicited charges that would ai agent do with the card :D
Kastra
@busmark_w_nika That's exactly the goal. AI agents are incredibly capable, but they can also make high-impact mistakes when given real autonomy. Kastra evaluates every action before it executes and enforces deterministic policies, so instead of discovering a destructive action after it happens, you can prevent it, require approval, or allow it based on your own rules. Think of it as an authorization layer for AI agents; you decide what they're allowed to do before they do it.
Governing agents before they act instead of auditing after the fact is the shift I've been hoping for.
Kastra
Kastra
@charlos_brat Exactly! And the nicest part is it isn't either/or. Every decision Kastra makes before an action runs is written to a tamper-evident, per-tenant hash-chained decision log, so the "audit after the fact" is just the other face of governing before it. And because it's chained, you can prove nothing was altered or dropped. There's a chain-verification API an auditor runs, not a log file you take on faith.
Macaly
runtime authz is exactly the missing peice for agents 🔥 how granular do the policies get?
Kastra
@petrkovacik Thanks! Pretty granular. Policies can evaluate prompts, models, users, tool calls, shell commands, file paths, Git operations, databases, APIs, environment, secrets, and even the arguments passed to each action. You can allow, deny, or require approval based on any combination of those. It's up to your creativity.
I like that this doesn't ask me to trust the agent's judgment at all, it just checks the action against a rule before anything happens. That's the right mental model. Most of the "AI safety" tooling I've evaluated this year still leans on the model behaving well, which never sat right with me.
Kastra
@thomas_jack3 Very accurate! We learned this by testing the product with some of the top enterprises pushing AI products into production. We realized autonomous AI products required a new infra layer and we built Kastra to solve the trust problem in a scalable way.
Earth.fm
Kastra
@1mirul exactly! The primary goal is to deterministically stop things before they happen. The monitoring and logs are a natural byproduct, since we're already intercepting every request anyway.
Kastra
@1mirul Thanks! That's exactly how we think about it. Monitoring is valuable, but once an action has already happened, you're often in incident response mode. We wanted to give developers a deterministic decision point before execution, so they can safely delegate more work to AI without giving up control.
Finally something that treats agent permissions like a real problem. The sub-1ms latency claim seems legit - I didn't notice any lag in tool calls when I tested it with Claude Code.
Kastra
Kastra
@baharyakupbcr2 Awesome Bahar, glad you had a good experience! Keeping the AI flow fast was a concern for us since day one. Every decision was made around adding this security layer without hurting performance.