Second-model AI auto-review for DeepSeek Harness approval requests: a read-only reviewer subagent returns structured allow/deny verdicts with reasons, fail-closed by default, fully auditable from the session log (approval/asked -> autoReview/verdict -> approval/decided).

Bounded, layered, approval-gated, auditable cross-session memory for DeepSeek Harness: a ctx.memory capability seam with a zero-dependency SQLite provider, a memory tool, and frozen snapshot injection. Maintenance status: FROZEN since 2026-10-05.

Your personal AI runtime, local-first. Patchwork OS gives any AI model a consistent set of tools, YAML recipes, a delegation policy with approval queue, and a durable trace memory — all on your machine, all under your policy.

Claude Code-style declarative permission rules for DeepSeek Harness: ordered allow/deny/ask rules with tool-name, argument (glob/regex), and workspace-path matching on the tools/pre-execute waterfall, session-log audit, and HMR reload.

Lark/Feishu real-time event listening / subscribing / consuming: stream events as NDJSON via `lark-cli event consume <EventKey>` (covers IM messages/reactions/chat changes, Approval status changes, Task updates, VC meeting started/joined/ended, Minutes generated, Whiteboard updated, etc.). Use for L

End-to-end orchestrator: from a business idea, app idea, or existing app to running Azure deployment with cost estimates and pre-deploy approval. Analyzes your app, auto-detects the right Azure services, scaffolds infrastructure code, and deploys — tailored to your app, not a template. Handles movin

MCP management console for the official DeepSeek Harness MCP client: /mcp command with health diagnostics and pipeline trial calls, a Settings MCP tab with server CRUD (approval-gated writes, automatic backups) and a tool trial console over the official tool pipeline (Apache-2.0, dsh-plugin).

MCP server for SSH: let Claude Code, Codex and any MCP client run commands, transfer files and manage servers over SSH, with human approval, an encrypted vault and a live audit trail. Desktop app and CLI included.

Approval gate and audit trail for AI agent actions: policy check first, proof after. Any MCP client.

A local MCP server that places a safety layer between AI coding agents and Postgres, parsing every statement, holding risky or large writes for human approval, and journaling all activity on your machine. It enables one-command undo of individual agent-made changes and blocks destructive operations outright.

A free, read-only server that lets users diagnose production AI-agent risks: it flags over-privileged API keys from permission descriptions, scans code snippets for the signals that prevent duplicate execution, and explains guard patterns such as idempotency keys, approval gates, outboxes, and circuit breakers. It makes no network calls or writes, and rejects inputs that look like real secrets.

Official-grade GitHub CI for DeepSeek Harness: composite action (review PRs / fix CI / report), polling PR review bot with idempotent inline comments, a status-check gate, and approval-gated PR/issue/repo/file tools

Evidence-backed UI delivery for coding agents working on existing Web products: declared criteria, criterion-bound artifacts, findings linked to declarations, and repair guidance. Evaluator review does not replace human semantic approval.

Your personal AI runtime, local-first. Patchwork OS gives any AI model a consistent set of tools, YAML recipes, a delegation policy with approval queue, and a durable trace memory — all on your machine, all under your policy.

Opt-in, token-authenticated MCP server built into the Shelly Android terminal app: read-only tools for terminal output, git status, registered agents and repos on the phone, plus off-by-default run_command and write_file that need an on-device approval tap per call.

Enables CLI agents to fold source documents into a cited Markdown knowledge base using an ontology, exposing tools to list, read, search, lint, and write pages only after user approval.

Enables MCP clients to create and control persistent, PTY-backed terminal sessions on macOS: launch commands and interactive programs, read output incrementally via byte cursors, send keystrokes, wait on conditions, and detect running/awaiting-input/exited states. Sessions run inside a Seatbelt sandbox with command policies, human approvals, secret redaction, and hash-chained audit logging.

Enables MCP clients such as ChatGPT to execute shell commands, manage files, and view or capture images on Mac and Linux hosts. Supports terminal-based connection approval, time-limited access, and output controls.

Enables MCP-capable agent CLIs on macOS to reuse an installed, unmodified OpenAI Browser Runtime to drive a connected Chrome or Edge browser (or a host's in-app browser backend) through js, js_reset, and turn_ended tools. It stays a thin bridge that preserves the Runtime's approval and security boundaries, failing closed when the client cannot handle approval requests.

Enables MCP clients such as Claude to query mail, schedules, and electronic approvals, reserve meeting rooms, and manage personal schedules through a groupware account.

Enables AI clients to inspect local RAM, commit, page file, disk usage, top processes, disk scans, duplicate files, startup apps, recent changes, and reports. It also generates approval-gated plans for cleanup, file moves, and startup disabling without changing the PC.

Connects to a logged-in Windows LINE Desktop client to read chat history and cached images for a given chat and date range, and to perform GUI-driven actions such as batch sending text, files, real @mentions, quote replies, and forwarding after a single approval. It also supports read-only cross-chat search over the last 31 days, recent-chat listing, and local receipts that stop rather than resend when the outcome is uncertain.

Audit trail and traceability for AI agents. See what every coding agent did, block risky calls, ask a human for approval, and prove it with signed trust checks.

Your pocket agent. Local-first AI agents on iOS and Android — real workspaces, tool execution with approvals, and your choice of model (DSH · Claude Code · Codex · GLM).