Releases·Tool Release
Show HN: Opair, a coding harness that eschews autonomy
A developer released Opair, a coding harness designed to limit agent autonomy on Hacker News. It replaces shell access with restricted, development-specific tools and gates file reads and writes behind access control. The release matters to practitioners building agentic coding workflows who want tighter sandboxing by default.
2 engagement1 sourceWed, Sep 9, 2026, 11:54 AM
AI Overview
- Opair restricts agents to development-specific tools, gates reads/writes by access control, and ships read-only MCP integrations with no git commit tools.
- Opair splits driver and navigator modes: navigator agents get no writable tools and monitor changes as in human pair programming.
Entities
Related
Products·building
RepoRelay: Secure MCP bridge for read-only AI code review
RepoRelay is an open-source MCP bridge that lets ChatGPT or Claude inspect a single approved local repository with read-only access, no shell, no Git, and no arbitrary filesystem access. It solves the problem of giving AI coding tools too much control over your machine, enabling safe code analysis and review.
RepoRelay is an open-source MCP bridge that lets ChatGPT or Claude inspect a single approved local repository with read-only access, no shell, no Git, and no arbitrary filesystem access. It solves the problem of giving AI coding tools too much control over your machine, enabling safe code analysis and review.
0 engagement1 sourceredditWed, Aug 19, 2026, 12:15 AM
Products·building
RunWield: AI coding harness with plan-before-edit workflow
RunWield is a coding harness that gives developers more control over AI agents by requiring a written plan before non-trivial code changes, and supporting ideation and prototyping before committing. It addresses the problem of reviewing large diffs after AI edits by making the agent's reasoning explicit upfront.
RunWield is a coding harness that gives developers more control over AI agents by requiring a written plan before non-trivial code changes, and supporting ideation and prototyping before committing. It addresses the problem of reviewing large diffs after AI edits by making the agent's reasoning explicit upfront.
1 engagement1 sourceredditMon, Aug 17, 2026, 04:08 PM
Products·launched
Addom – open-source, telemetry-free AI coding harness and editor
Addom is an open-source desktop application combining an AI coding harness with a built-in editor. The LLM is used via tool calling and streaming to build applications, with local artifacts for reverting file changes, a file editor, and a memory tab for past prompts. It targets developers who want a transparent, telemetry-free alternative to existing AI coding tools.
Addom is an open-source desktop application combining an AI coding harness with a built-in editor. The LLM is used via tool calling and streaming to build applications, with local artifacts for reverting file changes, a file editor, and a memory tab for past prompts. It targets developers who want a transparent, telemetry-free alternative to existing AI coding tools.
2 engagement1 sourcehackernewsThu, Sep 3, 2026, 06:18 PM
Products·launched
o8: Open-source orchestrator for managing multiple coding agents
o8 is an open-source control room for running multiple coding agents (Claude Code, Codex, Gemini, Aider, Goose, etc.) behind a single adapter contract. It scopes work, dispatches to agents in isolated git worktrees, and enforces a merge gate with rebase, typecheck, and head-SHA lock, requiring human approval before merging. It solves the problem of managing multiple agents without shared memory or audit trails, for developers using AI coding tools.
o8 is an open-source control room for running multiple coding agents (Claude Code, Codex, Gemini, Aider, Goose, etc.) behind a single adapter contract. It scopes work, dispatches to agents in isolated git worktrees, and enforces a merge gate with rebase, typecheck, and head-SHA lock, requiring human approval before merging. It solves the problem of managing multiple agents without shared memory or audit trails, for developers using AI coding tools.
23 engagement1 sourceredditSun, Aug 2, 2026, 03:16 PM
Products·building
Poly: Shared AI coding sessions with team review and approvals
Poly is a collaboration tool that turns AI coding sessions into shared, reviewable project rooms. It lets teams steer Claude Code together, with a shared timeline of prompts, replies, diffs, and cost receipts, and approvals from any screen. It solves the trust problem of AI-generated code by making it visible and attributable.
Poly is a collaboration tool that turns AI coding sessions into shared, reviewable project rooms. It lets teams steer Claude Code together, with a shared timeline of prompts, replies, diffs, and cost receipts, and approvals from any screen. It solves the trust problem of AI-generated code by making it visible and attributable.
2 engagement1 sourceredditTue, Aug 4, 2026, 01:35 PM