For Claude & any MCP client · free · private

Run the toolkit inside Claude.

Rate the five gears, compute the control ratio envelope for a fast system, evaluate the evidence with one veto and permit table, and pre-commit the brakes, all without leaving the conversation. Also available as an MCP server for any client that speaks the protocol. Specification 1.0.0.

Add the Claude Skill Get the Claude Code plugin Run the MCP server

Free · no accounts, no API keys · companion to the book by Abdul Kavungal (AK)

/equilibrium-score our new autonomous code agent
Equilibrium Scorecard · RSI-relevant system
Governance Equity Aligned incentives Resilience Steering
Advisory result: Blocked.
A red Resilience gear blocks the activity assessed, and an RSI-relevant system needs independent review. Add a tested rollback path and reassess.
Three ways in

Pick how you connect

Chatting in the Claude app? Add the Skill. Working in the terminal with Claude Code? Install the plugin. Using Cursor, VS Code, or another MCP client? Run the MCP server. Same policy module, same five gears, same one veto and permit table, all reading specification 1.0.0.

Claude Skill · for the Claude app

Effective Equilibrium

effective-equilibrium.skill

A Skill that helps Claude organize the evidence for one scoped activity, reach the advisory result the one veto and permit table supports, and say exactly what is missing or blocking. It never grants a permission; a named person does that separately.

  • Rate the five gears green, yellow, red or unknown for the activity actually assessed, in canonical order, never averaged.
  • Compute the control ratio envelope R = L / H from intervention latency and the oversight validity horizon, with the band read from R upper.
  • Draft required interventions as measurable If-Then records, tested on the exact version assessed.
  • Run a five-minute audit on anything that moves faster than the people watching it.
  • Apply the one veto and permit table, track control drift between revisions, and produce the specification's six-line disclosure.
Install
  1. Download the .skill file below.
  2. In Claude, open Settings → Capabilities → Skills and upload it.
  3. Ask: "assess this pipeline's limited deployment"
Download the Skill Skills docs
Claude Code plugin · for the terminal

Acceleration Paradox Toolkit

acceleration-paradox-toolkit.plugin · v1.1.0 · MIT

Six slash commands and four skills that turn the method into commands you can run while you build. The bundled offline page runs the same 1.0.0 policy module as the web assessment tool.

  • /equilibrium-score a named activity or a pasted doc against the one veto and permit table.
  • /oversight-half-life (kept for compatibility) computes the ratio envelope against the oversight validity horizon.
  • /circuit-breaker to draft required interventions with owners, deadlines and a tested version.
  • /equilibrium-audit for the five-minute gut check.
  • /control-drift and /control-disclosure, implementing the open specification.
Install
  1. Download the .plugin bundle below.
  2. In Claude Code, run /plugin and add it from the bundle.
  3. Type /equilibrium-score or ask in plain language.
Download the plugin Plugin docs
MCP server · for any MCP client

Acceleration Paradox MCP

mcp-server/ · local install, no npm registry

The website's canonical policy module, byte for byte, as tools, resources, and prompts over the Model Context Protocol, for Claude Desktop, Claude Code, Cursor, VS Code, or anything else that speaks MCP.

  • evaluate_assessment runs a full evidence pack through the one veto and permit table; compute_control_ratio returns the envelope with the band from R upper.
  • check_gear_ratings, get_example_assessment, track_control_drift and generate_disclosure, implementing the open specification.
  • Doctrine, gear anchors, the decision table, the evidence pack schema, the Harbor examples and the five-minute audit as readable resources.
  • Guided prompts: assess_activity, five_minute_audit, control_ratio_check, control_surface_review.
  • Runs locally over stdio and makes no network requests. Your MCP client's own data handling governs what reaches a model.
Install
  1. Download the source below and unzip it, then run npm install inside mcp-server/.
  2. Point your MCP client's config at node /path/to/mcp-server/index.js.
  3. Ask: "assess this canary deployment" or "what is our control ratio envelope?"
Download the MCP server MCP docs
In plain language

Just ask Claude

No syntax to learn. The Skill and the plugin both respond to natural prompts.

Assess it

"Assess the limited deployment of our new coding agent: gears, timing, hazards, tests."

Check timing

"Work out the control ratio envelope for this pipeline: the three L components against the oversight validity horizon."

Set guardrails

"Draft the required interventions for this rollout, with owners, deadlines and the version they were tested on."

Quick audit

"Run a five-minute risk audit on this launch plan."
What's inside the plugin

Four skills, six commands,
one offline page

TypeNamePurpose
Skillequilibrium-scorecardRate the five gears with the 1.0.0 anchors and apply the one veto and permit table to reach an advisory result, in any of the output formats.
Skilloversight-half-lifeCompute the control ratio envelope R = L / H against the oversight validity horizon (the skill keeps its historical name).
Skillcircuit-breaker-cardsDraft required interventions with owners, deadlines and a tested version.
Skillcontrol-surfaceTrack control drift between two 1.0.0 revisions of the same scope and render the specification's six-line disclosure.
Command/equilibrium-scoreAssess a named activity or a pasted document.
Command/oversight-half-lifeCompute the ratio envelope for one failure scenario.
Command/circuit-breakerDraft required-intervention cards for a system or risk.
Command/equilibrium-auditThe five-minute gut check.
Command/control-driftCompare R upper and the five gears with the previous revision; a rising R upper or any gear moving towards red is a regression.
Command/control-disclosureProduce the six-line control surface disclosure, with the advisory result and the human decision kept apart.
Pageequilibrium-scorecard.htmlThe offline assessment page: the same 1.0.0 policy module as the web assessment tool, with the Harbor and CALL-2026 examples built in. Nothing leaves the device.

Outputs: an advisory result in chat, a Word or PDF report, or a fillable markdown record.
This assessment organizes evidence for a scoped decision. Its ratio band is a policy indicator, not a safety certification or permission to operate.

What's inside the MCP server

Eight tools, seven resources,
four prompts

TypeNamePurpose
Toolcompute_control_ratioThe ratio envelope from planning bounds: L = detection + decision + execution delay over the oversight validity horizon H. Band from R upper: green below 0.25, amber to below 1, red at or above 1. Blanks stay unknown.
Toolevaluate_assessmentRun a complete evidence pack through the one veto and permit table: Blocked, Insufficient evidence, Eligible with conditions, or Eligible for scoped permission, with every reason. The human decision is echoed, never filled in.
Toolcheck_gear_ratingsGear-only pre-check in canonical order (green, yellow, red, unknown) plus independent-review routing from the five stakes judgments, rights and RSI relevance.
Toolget_example_assessmentThe fictional Harbor snapshots H1, H2 and H3 as complete records, evaluated at their recorded time.
Tooldraft_circuit_breakerA required-intervention record with trigger, action, owner, deadline, tested version and test result, and how the decision table reads it.
Toollist_circuit_breakersIllustrative interventions, one per gear, to start from.
Tooltrack_control_driftCompare two 1.0.0 revisions of the same scope: a rising R upper or any gear toward red is a regression. Legacy snapshots are reported as not comparable.
Toolgenerate_disclosureRender the specification's six-line disclosure, with the advisory result, the human decision and the version kept apart.
Resourceequilibrium://doctrineThe book's argument in 1.0.0 vocabulary: the two clocks, L, H, the envelope, the five gears.
Resourceequilibrium://gear-rubricThe 1.0.0 anchors for each gear and the evidence to show beside a rating.
Resourceequilibrium://five-minute-auditOne question per gear, in canonical order. A first read, never a result.
Resourceequilibrium://control-surface-specA summary of the Open Control Surface Specification 1.0.0 this server implements.
Resourceequilibrium://veto-rulesThe one veto and permit table: five precedence rows, most restrictive first.
Resourceequilibrium://evidence-pack-schemaThe Draft 2020-12 JSON Schema for evidence packs (1.0.0).
Resourceequilibrium://harbor-examplesThe Harbor H1/H2/H3 records plus the illustrative CALL-2026 preset, as JSON.
Promptassess_activityWalks the client through building and evaluating an evidence pack for one activity.
Promptfive_minute_auditAsks the five gear questions about a system, fast, then runs the gear pre-check.
Promptcontrol_ratio_checkWalks through the three L components and the remaining H with bounds, then computes the envelope.
Promptcontrol_surface_reviewRuns the four-step assessment and a drift check against a previous 1.0.0 revision.

The same policy module as the web assessment tool, byte for byte, passing the same 26 decision fixtures, and implementing the Open Control Surface Specification 1.0.0. Stdio transport, local only. This assessment organizes evidence for a scoped decision. Its ratio band is a policy indicator, not a safety certification or permission to operate.