HiddenLayer
Ai Security
Security platform for AI models and the ML lifecycle
ai-securitymodel-securityadversarial-ml
View →
HOL Guard
An open-source (Apache-2.0), local-first runtime control that intercepts risky AI coding-agent actions and lets you a...
Editorially reviewedChecked Oct 2026How we review
Free plan Free trial API Self-hosted Teams
Toolglade’s take
HOL Guard's open-source (Apache-2.0), local-first approval layer is a sensible 'checkpoint before the dangerous command' for AI coding agents, and local protection is free with no account. The vendor is unusually candid about limits: coverage is harness-specific and it is explicitly not a firewall or complete prompt-injection preventer. Download/star counts are self-reported.
Should you use HOL Guard?
- Best for
- developers running autonomous coding agents · teams wanting shared runtime policy
- Not ideal for
- users expecting complete prompt-injection protection · unsupported agents
- Biggest strength
- Core runtime is open source (Apache-2.0), free, local
- Biggest weakness
- Coverage is harness- and event-specific
- Pricing
- Free plan, paid from $4.99/mo
- Top alternative
- HiddenLayer · see all
Facts & pricing last checked October 2026. How we review
About HOL Guard
An open-source, local-first policy layer that can allow, ask, or block risky AI coding-agent actions before they run, plus a plugin trust scanner.
HOL Guard is a local-first runtime security layer for AI coding agents. Its core Guard Local runtime is open source under Apache-2.0 and runs on the developer's machine without a cloud account: it evaluates agent actions such as shell commands, secret access, MCP configuration changes, and plugin installs, and can allow, ask, or block them before they run. The vendor is explicit that coverage is harness- and event-specific and that Guard is not a network firewall or a complete prompt-injection preventer. It supports a growing set of agents (Codex, Claude Code, Cursor, OpenCode, and others) and includes a plugin scanner that assigns trust scores and can gate pull requests in CI. An optional Guard Cloud adds synced decision history, alerts, and shared team policy.
Weighing your options?See how HOL Guard compares to the alternatives.
TL;DR
An open-source (Apache-2.0), local-first runtime control that intercepts risky AI coding-agent actions and lets you allow, ask, or block them, with optional paid cloud sync.
Company overview
Published by HOL (Hashgraph Online), a consortium building agentic-AI standards (footer: HOL DAO LLC). Developed in the open on GitHub (hashgraph-online/hol-guard).
Product features
Local allow/ask/block enforcement, local audit receipts, multi-harness support, a plugin scanner with trust scores and a CI Action, and optional Guard Cloud for sync and team policy.
Target market
Developers and teams running AI coding agents who want a runtime safety checkpoint, plus plugin maintainers needing pre-release gating.
Buyer personas
End users
Developers using AI coding agents who want to stop destructive commands and secret leaks.
Buyers
Engineering/security leads standardizing policy across a team.
Key influencers
Security engineers and open-source maintainers.
Ideal customer profile
Security-conscious teams running autonomous coding agents who want open, local-first controls with optional central policy.
Funding & performance
Not disclosed; solicits sponsorships.
Pros & cons
Pros
- Core runtime is open source (Apache-2.0), free, local
- Allow/ask/block before an action has side effects
- Works across many coding agents
- Vendor is candid about coverage limits
Cons
- Coverage is harness- and event-specific
- Explicitly not a firewall or complete prompt-injection preventer
- Many headline metrics are self-reported
- Strongest team value sits behind paid cloud
Pricing plans
Free (Guard Local)
$0
- Open-source Apache-2.0 local runtime
- Allow/ask/block on supported actions
- Local secret scanning & audit receipts
- Full offline support, no account
Solo
$4.99 / month
- Cross-device decision memory
- GitHub monitoring of up to 5 repos
- 250 credential checks/month
Pro
$15 / month
- 5 cloud devices
- 180-day retention, alerts
- Policy versioning and full evidence
Team
Not listed / month
- Organization/shared policy
- Team workflows and dashboards
Enterprise
Contact sales
- Self-hosted / VPC review
- Custom threat feeds
- Volume pricing
Key features
API
Team collaboration
Self-hosted
Integrations
Claude Code, Codex, Cursor, OpenCode, GitHub Actions
Input types
Agent shell commands, Tool/MCP calls, Plugin installs, Config changes
Output types
Allow/ask/block decisions, Local audit receipts, Trust scores
Best For
AI coding agent users, Security-conscious developers, Plugin maintainers
Compare key features
View all alternatives →Feature
HOL Guard
HiddenLayer
Kilo Code
Pricing
Freemium
Paid
Freemium
Free plan
Yes
No
Yes
Free trial
Yes
No
No
API
Yes
Yes
Yes
Self-hosted
Yes
Yes
Yes
Team support
Yes
Yes
Yes
Frequently asked questions
Is HOL Guard open source?+
Yes. The core Guard Local runtime is Apache-2.0 and usable without a paid plan or account.
Does it work offline?+
Yes. Guard Local evaluates supported actions on your machine; local protection needs no account or internet.
Is it a complete prompt-injection solution?+
No. The vendor states it reduces impact at supported boundaries but is not a complete prompt-injection preventer.
Reviews
Loading reviews…
Compare
Compare HOL Guard with other AI tools
Side-by-side pages for pricing, features, and best-fit use cases.
Compare side by side
HOL Guard vs HiddenLayer
Compare →
Compare side by side
HOL Guard vs Kilo Code
Compare →
Compare side by side
HOL Guard vs Glama
Compare →
Compare side by side
HOL Guard vs OpenObserve
Compare →
Compare side by side
HOL Guard vs FastMCP
Compare →
Compare side by side
HOL Guard vs GoModel
Compare →
Similar tools you may like
Paid
Kilo Code
Coding
Open-source, model-agnostic AI coding agent for VS Code and JetBrains
ai-coding-agentopen-sourcevs-code
Free planView →
Glama
MCP
MCP server registry, in-browser inspector, and gateway with an LLM API bundled in.
mcpregistrygateway
Free planView →
OpenObserve
Llm Observability
OpenTelemetry-native observability for LLM calls, tools, and agent handoffs
opentelemetryllm tracingagent observability
Free planView →
Learn to use HOL Guard well
Free, hands-on Toolglade courses that cover this kind of tool — start reading in a minute, no sign-up needed.
AI Safety & Red-Teaming
Test, harden, and govern LLM systems against misuse and failure — prompt injection, jailbreaks, agentic risks, and the defense-in-depth reality that there is no single fix.
Module 1 freeAdvanced~5h
Secure AI-Assisted Development
Ship faster with AI coding tools without shipping vulnerabilities, leaked secrets, or risky dependencies.
Free courseIntermediate~4h