Repository navigation
Home
OpenSocrates helps the active agent produce results a person can understand, judge, and use. Version 1.5.0 preserves 48 authored reasoning methods and all 96 English/Korean procedure bodies. Supplemental reader guidance connects conclusions to their sources, stakeholder context, and useful next actions.
The release supports Codex on Apple-silicon Mac and Windows x64, native Claude Code on Apple-silicon Mac, portable Claude account skills, and Antigravity workspace/global content on Mac and Windows x64. Dedicated coding-specialist, project-memory, and broad orchestration features are outside v1.5.
The Wiki provides orientation. Tagged release documentation, versioned source, SECURITY.md, CONTRIBUTING.md, and AGENTS.md remain authoritative. Implementation, installation, live delivery, application, and useful outcomes are distinct.
- Install or update: Host-Support-and-Installation
- Understand delivery and grounding: Architecture
- Interpret verification claims: Validation-and-Evidence
- Contribute and release: Development-and-Release-Workflow
- See priorities and historical direction: Roadmap
- Canonical overview: README
- Release files: v1.5.0
- Report a vulnerability: SECURITY.md
The active agent retrieves the smallest sufficient eligible method set when a judgment materially changes. It reads complete canonical procedures, checks their applicability and stop conditions, and preserves the user's goals, permissions, constraints, and requested output. Mechanical work needs no added reasoning ceremony. A failed or unavailable entry does not block ordinary work.
Reader guidance helps reconcile dated sources, preserve settled facts and unknowns, explain differing stakeholder perspectives, and write practical messages without inventing motives, owners, or commitments. It supplements the authored procedures rather than rewriting them.
After an eligible canonical method is read in full and actually applied, the
final answer uses the exact line Powered by OpenSocrates, unchanged across
languages. Reader guidance alone does not qualify. The footer is attribution,
not independent application evidence or proof of a better result.
- No OpenSocrates backend, telemetry, separate account, or API key.
- No raw prompts, transcripts, credentials, screenshots, workspace content, or hidden reasoning in OpenSocrates records.
- Host login, permissions, conversation storage, and service terms still apply.
- Canonical Method IDs, revisions, and content identities remain available in the current context and existing permitted evidence; the footer adds no logging.
- No general quality, token-cost, or latency improvement is claimed.
See the Mac guide, Windows guide, and changelog for version-specific details. Bug reports should describe the observed failure and environment without private prompts, transcripts, or unsanitized logs.