OneCLI solves a problem that has quietly haunted every team deploying AI agents in production: how do you give an agent access to infrastructure without giving it the keys to the kingdom?

On July 23, 2026, the open-source project OneCLI hit the front page of Hacker News with 6400 upvotes and 40% growth velocity in community mentions. The signal was clear: practitioners were tired of the trade-off between automation and security.

This article explains why OneCLI matters, how it works, and how to integrate it into your automation stack without breaking existing workflows.

The Core Question

How do you run 'show' commands across your infrastructure, databases, and network devices without embedding credentials in scripts, environment files, or – worst of all – AI agent prompts?

The tension is real. Every team I've worked with has faced it. You want real-time visibility. You want AI agents that can diagnose issues, generate reports, and trigger alerts. But the moment an agent has access to a credential, that credential is one prompt injection away from exfiltration.

OneCLI answers this question with a simple architectural pattern: a local gateway that sits between your agent and your infrastructure, injecting credentials at runtime without the agent ever seeing them.

What Most People Get Wrong

Most teams approach credential management for AI agents by doing one of three things:

1. Embedding credentials in agent prompts – A pattern so dangerous it should be illegal. According to the OWASP Top 10 for LLM Applications 2025, prompt injection attacks increased by 340% year-over-year, with credential exfiltration as the primary payload.

2. Using environment variables – Slightly better, but still exposes credentials to any process that can read /proc or dump environment memory. A compromised agent container leaks everything.

3. Hardcoding in scripts – The worst of all. Git history becomes a credential dump.

What these approaches miss is that the agent doesn't need the credential. It needs the output. OneCLI inverts the model: the agent issues a 'show' command, the gateway authenticates on its behalf, and returns only the result.

The Expert Take

OneCLI is not a new idea. It's a well-executed implementation of a pattern that security-conscious teams have been building internally for years. What makes it notable is its open-source nature, its simplicity, and its integration hooks for AI agents.

Let me walk through the architecture.

How OneCLI Works

OneCLI operates as a local proxy. You install it on a bastion host, a jump box, or directly on the target machine. It exposes a Unix socket or a local HTTP endpoint. Your AI agent connects to this endpoint and issues commands.

The flow looks like this:

1. Agent sends a command (e.g., show running-config)
2. OneCLI intercepts the command
3. OneCLI looks up the target device in its credential store (local encrypted file, HashiCorp Vault, or environment variable)
4. OneCLI authenticates to the target device
5. OneCLI executes the command
6. OneCLI returns the output to the agent

The agent never sees the username, password, API key, or token.

Supported Targets

As of version 0.4.2 (released July 2026), OneCLI supports:

• Network devices: Cisco IOS, Juniper JunOS, Arista EOS
• Databases: PostgreSQL, MySQL, MongoDB
• Cloud APIs: AWS CLI, Azure CLI, GCP gcloud
• Custom commands: Any command that can be wrapped in a shell script

Security Model

OneCLI's security model is straightforward:

• Credentials are stored encrypted at rest using AES-256-GCM
• The gateway process runs as a non-root user
• Commands are logged with timestamps and agent identifiers
• Rate limiting prevents brute-force attacks
• Audit trail captures every command execution

According to the project's threat model documentation, the primary attack vector is a compromised agent that sends malicious commands. OneCLI mitigates this by allowing administrators to define command whitelists and blacklists.

Supporting Evidence & Examples

Real-World Case Study: AcmeCorp Infrastructure Monitoring

In Q2 2026, Maria Chen, a platform engineer at a 200-person SaaS company, was spending 6 hours per week manually running 'show' commands across 47 network devices to verify configuration compliance. She had tried automating with Ansible, but the playbooks contained hardcoded credentials that the security team flagged during audit.

Maria deployed OneCLI on a bastion host and connected it to a custom AI agent built on Claude 3.5 Sonnet. The agent was configured to run a daily compliance check: it would issue show running-config on each device, compare the output against a baseline, and flag any drift.

Result: Compliance checks dropped from 6 hours to 12 minutes per week. Zero credentials exposed. The security team approved the workflow because the agent had no direct access to any credential store.

Comparison: Manual vs. Automated 'show' Command Execution

Nuances Worth Knowing

OneCLI Is Not a Panacea

OneCLI solves credential exposure, but it introduces its own attack surface. If an attacker compromises the OneCLI gateway itself, they gain access to all stored credentials. The project recommends running OneCLI on a dedicated, hardened host with minimal services.

Command Whitelisting Is Essential

Without whitelisting, an agent could issue destructive commands like delete running-config or DROP DATABASE. OneCLI supports regex-based whitelists. For example:

whitelist:
  - "^show .*"
  - "^ping .*"
  - "^traceroute .*"

This limits the agent to read-only and diagnostic commands.

Integration with Workflow Automation Platforms

OneCLI's local HTTP endpoint makes it easy to integrate with no-code automation platforms. For example, in Make.com, you can use the HTTP module to send a POST request to http://localhost:8080/execute with a JSON body containing the command and target device.

Here's a step-by-step guide to setting up a monitoring workflow:

1. Install OneCLI on a bastion host using the official installer: curl -sSL https://get.onecli.dev | bash
2. Configure credentials by running onecli add target --name core-router-01 --type cisco --host 10.0.1.1
3. Start the gateway with onecli serve --port 8080
4. Create a Make.com scenario with a scheduled trigger (e.g., every 6 hours)
5. Add an HTTP module that sends a POST to http://bastion-host:8080/execute with body {"target": "core-router-01", "command": "show running-config"}
6. Parse the response using Make.com's JSON parser
7. Add a filter to check for configuration drift
8. Add a notification module (Slack, email) if drift is detected

This entire workflow can be deployed from Neura Market's automation marketplace in under 30 minutes.

Latency Considerations

OneCLI adds approximately 50-200ms per command due to authentication and encryption overhead. For most monitoring use cases, this is negligible. For real-time interactive use, consider running the gateway on the same host as the agent to minimize network latency.

Practical Implications

For Security Teams

OneCLI provides a clear audit trail and eliminates credential sharing. This directly addresses SOC 2, ISO 27001, and PCI DSS requirements around access control and credential management.

For Platform Engineers

You can now give AI agents read-only access to infrastructure without the risk of credential leakage. This enables automated incident response, configuration compliance, and capacity planning without security review bottlenecks.

For Business Decision-Makers

The cost savings are measurable. According to a 2026 report by Forrester Research, organizations that automate infrastructure monitoring reduce mean time to resolution (MTTR) by 62% and save an average of $1.2 million annually in operational costs. OneCLI makes this automation secure enough to pass any audit.

Looking Ahead

OneCLI's roadmap includes:

• Dynamic credential rotation: Automatically rotate credentials after each use
• Multi-factor authentication support: Integrate with TOTP and WebAuthn
• Federated identity: Support for OAuth 2.0 and OIDC
• Cloud-native deployment: Helm charts for Kubernetes, Terraform modules

As AI agents become more autonomous, the need for secure credential gateways will only grow. OneCLI represents a foundational pattern: the agent talks to the gateway, the gateway talks to the infrastructure, and secrets never cross the boundary.

Summary & Recommendations

OneCLI is not just another open-source tool. It's a necessary architectural pattern for any organization deploying AI agents in production. The core insight is simple: agents don't need credentials, they need command outputs.

Recommendations:

1. Deploy OneCLI on a dedicated bastion host for all infrastructure monitoring use cases
2. Define strict command whitelists to limit agents to read-only operations
3. Integrate with your automation platform using the HTTP API
4. Monitor OneCLI logs for anomalous command patterns
5. Audit your existing workflows for credential exposure and migrate to OneCLI

For teams looking to accelerate their automation journey, browse the AI agent templates on Neura Market that integrate with OneCLI out of the box.

FAQ

Q: Is OneCLI production-ready? A: Yes, version 0.4.2 is stable and has been tested in production environments with up to 500 concurrent agents.

Q: Does OneCLI support Windows? A: Not yet. The project targets Linux and macOS. Windows support is on the roadmap for Q4 2026.

Q: Can I use OneCLI with cloud APIs like AWS? A: Yes. OneCLI can execute AWS CLI commands with temporary credentials generated via STS.

Q: How does OneCLI handle credential rotation? A: Currently, credentials are static until manually updated. Dynamic rotation is on the roadmap.

Q: What happens if the OneCLI gateway goes down? A: Agents lose access to infrastructure. We recommend running OneCLI in a high-availability configuration with a load balancer.

Q: Is OneCLI compatible with Zapier? A: Yes, via the Webhooks by Zapier app. Send a POST request to the OneCLI HTTP endpoint.

Q: Where can I find pre-built workflows for OneCLI? A: Check the Neura Market workflow directory for templates that include OneCLI integration.

OneCLI is available now at onecli.dev. For teams that want to go further, explore the full automation ecosystem on Neura Market.

Frequently Asked Questions

What is the best way to get started with Show HN: OneCLI – OSS Credential Gateway?

The best approach is to start with a clear goal in mind. Identify the specific workflow or process you want to automate, then explore the relevant templates and tools available on Neura Market to find a solution that matches your requirements.

How much does workflow automation typically cost?

Costs vary significantly depending on the platform and scale. Many automation platforms offer free tiers for basic workflows, with paid plans starting around $20–$50/month for small teams. Enterprise solutions can range from $500 to several thousand dollars per month. Neura Market offers templates for all major platforms so you can compare costs before committing.

Do I need technical skills to implement workflow automation?

Modern no-code and low-code platforms like Zapier, Make.com, and others have made automation accessible to non-technical users. Most workflows can be built using visual drag-and-drop interfaces without writing any code. For more complex integrations involving custom APIs or data transformations, some technical knowledge is helpful but not required for the majority of use cases.