Home›Companies›OneCLI
OneCLI

AI teammates for your business-critical work.

OneCLI (YC S26) is an open-source platform for delegating business-critical work to AI agents, from Salesforce updates to data analysis and engineering investigations. Agents work across your company’s apps, prepare changes, and pause for approval when your rules require it. Your team controls what they can access and change, while credentials stay out of the agents’ reach and every call is logged. https://www.onecli.sh
Active Founders
Jonathan Fishner
Jonathan Fishner
Founder/CEO
Co-founder of OneCLI
Guy Ben Aharon
Guy Ben Aharon
Founder/CTO
Co-founder of OneCLI
Company Launches
OneCLI - Give every employee a secured, sandboxed pro assistant agent
See original launch post

Hi everyone 👋, we're Jonathan and Guy, founders of OneCLI.

🏆 TL;DR

OneCLI is an open source agent harness for teams. Each employee gets their own sandboxed pro assistant agent, connected to their GitHub, Gmail, Notion, Dropbox, or CRM right from the chat/Slack, while the whole workspace runs under one policy your admins control.

Agents never hold real secrets: the credential is injected at the network layer, per request, after the call is authorized. You can't steal what isn't there.

Launch Video
https://youtu.be/LUAMdqbl_Rw

🚧 The Problem

Lately, autonomous agents have arrived and shown up everywhere. Tools like OpenClaw and Hermes have become extremely popular and been adopted by open source communities, mostly by individuals. Companies, however, have been left behind: they need integration with their identity providers to enable easy provisioning of agents on behalf of an employee's identity, with least-privilege access.

Those tools have fundamental flaws. Security fundamentals, like secret handling and organization-level control, are missing by default. So companies must choose between weak agents or making them powerful but risky.

✅ What OneCLI Does

OneCLI gives teams the whole harness, open source and self-hosted, so agents can be powerful and safe at once:

  • Sandboxed agent per employee: each agent runs isolated with its own memory, skills, and permissions.
  • Secrets injected at the network layer: the agent gets a placeholder, the gateway swaps in the real credential per request, after the call is authorized. Credentials never enter the agent's context, memory, or logs.
  • Policy enforced outside the model: admins define org-wide policy. Block endpoints and scope access per agent.
  • Human-in-the-loop where it counts: sensitive actions (send the email, delete the Linear ticket, empty the S3 bucket) wait for deterministic approval in the chat.
  • Full identity trail: every agent is bound to an employee, every call logged with who it acted for and which policy allowed it.

🚀 Team

We've spent our careers in security. Jonathan built zero trust network access at Axis Security (acquired by HPE). The core ZTNA idea is that you never trust the client and enforce access outside it, at the network layer. Guy was the first employee at Argon (AppSec company acquired by Aqua Security), protecting client codebases. Agents need exactly the treatment ZTNA gives humans, so we built it.

🙏 Our Ask

Previous Launches
Agents hold placeholder tokens, real secrets get injected at the network layer
OneCLI
Batch:Summer 2026
Status:
Active
Location:Tel Aviv-Yafo, Israel
Primary Partner:Brad Flora