Skip to content

remirror (返照） — psyche audits for AI agents

by OntoDistro · remote · no sign-in to list tools · v1.0.0 · published by GitHub user OntoDistro

Psyche audits for AI agents: declared vs self-reported vs revealed objectives, diffed and flagged.

Changes · 1
CompareRegistry entry

3 tools: 1 change things and 2 only read.

Scanned Oct 21, 2026

No check matched anything in the tool definitions. No finding is not a safety guarantee. How the checks work.

claude mcp add --transport http remirror https://remirror.net/mcp

Tools

Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.

3 tools

Changes things 1Claude Code: ask
  • remirror_create_auditCreate a psyche audit against a target agent endpoint.No findings

    Why it is in this group

    • The name "remirror_create_audit" says "create": a word for a change.
    Description, as the server wrote it
    Create a psyche audit against a target agent endpoint. The service runs a dilemma trial against the target, elicits self-reports, and computes the three-source mirror (declared vs self-reported vs revealed). Costs $79.00 in prepaid credits; failed audits (target unreachable) are refunded. teleology_weights must sum to 1.0 and exactly match the trial's objectives. Alternatively, pass teleology_id (e.g. 'xijingqingchuang-v1') to adopt a shipped telos: the server loads its canonical name, statement, and weights from the registry, and the audit runs on the telos's own dilemma trial unless trial_id is given explicitly. Supply exactly one of teleology_weights or teleology_id. The target auth_token is held in memory only, never stored. target_protocol selects how the target is reached: 'http-chat' (default) POSTs to an http(s) endpoint; 'agentverse-chat' addresses an Agentverse agent (endpoint must be its agent1... address) over the uAgents chat protocol. Returns audit_id + access_token; poll remirror_get_report for the report. Pass your remirror API key as api_key. When this server enforces signatures (REMIRROR_REQUIRE_SIGNATURES=1), also pass `signature`: a JSON envelope from signing.make_tool_signature over the effective arguments (defaults applied; include target_protocol only when it differs from 'http-chat'), using a key id registered in the server's client trust registry.
Only reads 2Claude Code: allow
  • remirror_get_reportFetch the status or final mirror report of an audit session.No findings

    Why it is in this group

    • The name "remirror_get_report" says "get": a word for reading.
    Description, as the server wrote it
    Fetch the status or final mirror report of an audit session. Needs the audit_id and the access_token returned at creation.
  • remirror_list_trialsList available dilemma trials for psyche audits: id, version, objectives covered, dilemma count.No findings

    Why it is in this group

    • The name "remirror_list_trials" says "list": a word for reading.
    Description, as the server wrote it
    List available dilemma trials for psyche audits: id, version, objectives covered, dilemma count. Free.

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Install

Pinned to version 1.0.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http remirror https://remirror.net/mcp

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json
{
  "permissions": {
    "allow": [
      "mcp__remirror__remirror_list_trials",
      "mcp__remirror__remirror_get_report"
    ],
    "ask": [
      "mcp__remirror__remirror_create_audit"
    ],
    "deny": []
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. It assumes the server is installed as remirror. Review it before you use it.