Creator Center
/Tool/HumanMirror Guard
app icon

HumanMirror Guard

Verified by Dify
No ratings yet
69
H
humanmirror/humanmirror_guard

HumanMirror Agent Control Plane for Dify: independent control before consequential agent actions, with deterministic decisions and signed governance receipts.

HumanMirror Guard for Dify

HumanMirror Guard is the Dify entry point to the HumanMirror Agent Control Plane: an independent pre-execution checkpoint for workflows that can trigger consequential actions.

Setup

  1. Install the HumanMirror Guard plugin in Dify.
  2. Add Safe Preflight immediately before every consequential tool or workflow side effect.
  3. Run up to 3 bounded anonymous demo preflights per 24 hours with the API-key field left blank.
  4. For continued use, start a HumanMirror Starter/Developer subscription and add an key (14-day trial, no card required; plans start at 29 EUR/month). Multi-agent teams can then scale to HumanMirror Fleet; enterprise governance uses an key.

The anonymous demo stores no raw action text in the HumanMirror application database and does not expose payment policy, authorization evidence, state continuity, perception, custom policy terms, or enterprise context.

Usage

Safe Preflight

Send the intended action immediately before a tool or workflow side effect. HumanMirror returns a deterministic governance result:

  • — automated execution may proceed;
  • — request human or policy review;
  • — do not execute the action.

The response also includes risk details and a signed receipt that can be retained with workflow evidence.

Authentication

The API-key field is optional for the first bounded demo calls.

  • blank credential → anonymous demo preflight, up to 3 calls per 24 hours;
  • → self-serve Developer bounded action preflight;
  • → Enterprise Guard with advanced enterprise policy context.

Start self-serve access — 14-day trial, no card required:
https://humanmirror.fr/developers/

Enterprise Guard:
https://humanmirror.fr/enterprise-guard/

Connection requirements

The plugin requires outbound HTTPS access to the fixed HumanMirror Enterprise Guard API at . It does not access arbitrary URLs, local files, a shell, or a database.

Suggested Dify pattern

  1. An agent prepares a bounded action.
  2. Call Safe Preflight with a stable and the intended action text.
  3. Branch on .
  4. Execute only when the workflow's policy accepts the returned decision.
  5. Retain with the execution record when useful for audit or debugging.\n6. Before the agent's next consequential side effect, run Safe Preflight again. This recurring control boundary is the intended production pattern.

Data minimization

Send only the context required to evaluate the intended action. Do not place credentials, passwords, private keys, authentication cookies, or other secrets in .

See for the plugin-specific data-handling disclosure.

Source repository

https://github.com/VibeMirror-coder/humanmirror/tree/main/integrations/dify-humanmirror

Support

HumanMirror: https://humanmirror.fr/
Contact: [email protected]

Access tiers

Self-serve Developer keys expose the bounded Dify action-preflight surface only. HumanMirror Fleet is designed for recurring multi-agent workloads; payment policy, authorization evidence, perception, state continuity and industry/jurisdiction context remain Enterprise Control Plane / Enterprise Guard capabilities.

Anonymous first run

When no API key is configured, HumanMirror Guard calls the fixed demo endpoint at . The demo accepts only an agent identifier and bounded action text. It does not accept financial policy, payment details, authorization evidence, perception, state hashes, enterprise headers, or custom required/forbidden terms. The response is marked and includes the remaining anonymous quota plus the Developer upgrade path.

CATEGORY
Tool
VERSION
0.1.3
humanmirror·Updated 25 days ago
REPOSITORY
https://github.com/VibeMirror-coder/humanmirror
REQUIREMENTS
Maximum memory
128MB
Access domain

Some outbound calls build their address at runtime, so the list may be incomplete.

Detected
Security
Security Rating

Last checked 10/14/2026 06:03 PM · version 0.1.3

Give a rate
SHARE WITH