OpenSocratesGitHub
Open source · v1.3.1 · Local integration

Give your AI a method
before an answer.

OpenSocrates brings 48 authored reasoning methods to the agents you already use. Examine assumptions, compare options and revisit a decision when the evidence changes—all within your task.

Install v1.3.1 Read the source
  • No API key
  • No OpenSocrates backend
  • No telemetry
Works with
ILLUSTRATIVE EXAMPLE DECISION 01
YOUR PROMPT

Which vendor fits our budget and accessibility requirements?

EXAMPLE METHODS
Trade-off AnalysisAssumption Mapping
02complete procedures for this decision
Compare with evidence

Illustration, not a live result. A new audit can trigger another decision point.

48Authored reasoning systems
06Host integrations
96English / Korean procedures
MITOpen-source license
01 — WHAT IT HELPS YOU CHECK

Assumptions. Options.
Evidence that matters.

Use an authored method to structure a comparison, examine a claim or identify what would change a recommendation.

A

Check the assumptions

Make assumptions and evidence limits explicit before committing to a conclusion. Keep the strength of the answer tied to the available support.

B

Read complete methods

Forty-eight authored methods cover decisions, evidence, causality, product thinking and strategy. Each includes when to use it and when to stop.

C

Reconsider when facts change

A new audit or changed constraint can reopen a decision in the same task. Straightforward factual and mechanical steps need no added method.

02 — HOW IT WORKS

One task.
Room to reconsider.

01

Ask normally

Keep working in your supported agent. Use its OpenSocrates entry point where explicit invocation is required.

02

Notice the decision

When the objective, options or evidence changes, the active agent can revisit the judgment within the same request.

03

Read the right method

Retrieve eligible, complete procedures with their examples and limits. Claude and Codex use the packaged native selector first when available.

04

Keep the result grounded

Keep each question’s conclusion, missing evidence and reopening conditions distinct. Ordinary work continues if the integration is unavailable.

03 — THE CATALOG

48 ways to
think on purpose.

Each system includes complete theory, a step-by-step procedure, evidence rules, stop conditions and examples in English and Korean. v1.3.1 retains the existing authored procedure bodies.

Explore the authored catalog
AbductionBayesian UpdatingCausal ReasoningCritical ThinkingDecision Tree AnalysisDesign ThinkingEvidence HierarchyFirst PrinciplesGame TheoryJobs to Be DoneLogic TreeMECEPremortem AnalysisScenario PlanningSocratic QuestioningSystems ThinkingTrade-off AnalysisValue of Information+30 more
04 — HOST SUPPORT

One reasoning layer.
Six host integrations.

OpenSocrates uses the host you already work in and your existing login. Support levels stay explicit so implementation is never mistaken for validation.

C
Code / Cowork plugin

Claude

  • Entry/opensocrates:opensocrates

Discovery where plugin hooks run; native retrieval when available.

Support details ↗
X
CLI / Desktop plugin

Codex

  • Entryopensocrates controller

Approve hooks once, then start a new task. Native selection first.

Support details ↗
O
Local same-turn bridge

OpenCode

  • EntryNative skill fallback

A local bridge adds one complete procedure to the current turn.

Support details ↗
G
Native content skill

Grok Build

  • Entry/opensocrates

Host-native skill selection or explicit invocation; no OpenSocrates hooks.

Support details ↗
C
Agent Plugin

Cursor

  • EntrySkill discovery / invocation

Content-only integration with complete internal method references.

Support details ↗
A
Content skill

Google Antigravity

  • EntryExplicit opensocrates skill

Explicit invocation; no hook, native runtime or extra selector model.

Support details ↗

Distributed native runtimes target Apple-silicon macOS. Package verification is separate from live host validation. Claude web and Desktop Chat use a separate standalone skill ZIP: export integrity is validated; cloud activation remains unvalidated. Chat setup and support limits ↗

05 — PRIVACY & CONTROL

Local by design.
Bounded by default.

01

No OpenSocrates backend

The integration runs locally. Model requests still use your chosen host service under its terms.

02

No API key to manage

Use your host’s existing authentication and provider configuration. No separate OpenSocrates account or key is needed.

03

No product telemetry

OpenSocrates adds no product telemetry or hosted account layer. See the security policy for local diagnostic boundaries.

04

Fail-open behavior

When an integration or method lookup is unavailable, ordinary work can continue with the task’s constraints intact.

06 — INSTALL

Install once.
Work in your agent.

Requires Node.js 20 or later and an authenticated host. Current release target: Apple-silicon macOS.

FOR

All ready hosts

v1.3.1
npx --yes opensocrates@1.3.1 install --host all

Detect supported, authenticated host CLIs and install for those that are ready.

CHOOSE ONE

Codex

v1.3.1
npx --yes opensocrates@1.3.1 install --host codex

Or set --host to claude, opencode, grok, cursor or antigravity.

CHOOSE ONE

Claude

v1.3.1
npx --yes opensocrates@1.3.1 install --host claude

Install the Claude Code / Cowork plugin. Web and Desktop Chat use the separate skill ZIP.

THEN

Open a new task after installation. In Codex, approve the OpenSocrates hooks in an interactive session. Use explicit invocation where your host requires it.

OPEN SOURCE · MIT LICENSED

Check the reasoning.
Keep the evidence in view.

Install OpenSocratesStar on GitHub ↗

Guided writing is not a guarantee of better answers. Synthetic tests include slow completion and repetitive prose; no general quality, cost or speed improvement is claimed. Release evidence · Patch release verification · 한국어 안내