SkillAgentSearch skills...

gatherthread

GatherThread 共序:本地 AI 编程 Agent 多人联机协作、实时共享上下文。Self-hostable collaboration for local AI coding agents, Codex Desktop connector & DeepSeek Harness plugin.

Install / Use

claude mcp add TH060419 -- npx -y github:TH060419/gatherthread

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

76/100

Supported Platforms

Claude Code
Claude Desktop
OpenAI Codex

Our assessment of gatherthread

gatherthread scores 76/100 on our quality scale, 3682nd of 4,597 Development & Engineering skills we index.

Its MCP Server is 11 KB long, well organised into 12 sections and no code examples: a thorough specification that gives an agent plenty to work with.

It has 10 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
29/30
Structure
13/20
Description
15/15
Adoption
4/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated yesterday, so gatherthread is actively maintained.
  • It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

gatherthread compared with similar skills

All 4 of these similar skills score higher than gatherthread; compare them before choosing.

SkillScoreStarsUpdatedFormat
gatherthread (this skill)by TH06041976101d agoMCP Server
Agent-Reachby Panniantong10099.4k4d agoCLAUDE.md
taste-skillby Leonxlnx10099.1ktodayCLAUDE.md
headroomby headroomlabs-ai10076.0k1d agoCLAUDE.md
ai-job-searchby MadsLorentzen10055.5ktodayCLAUDE.md

Frequently asked questions

How do I install gatherthread?
Run claude mcp add TH060419 -- npx -y github:TH060419/gatherthread. The install tabs above show the steps for each supported agent.
Which AI agents does gatherthread work with?
It is written for Claude Code, Claude Desktop and OpenAI Codex, as a MCP Server file. Other agents that read the same format can often use it too.
Is gatherthread safe to use?
It is Apache-2.0-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is gatherthread still maintained?
The repository was last updated yesterday, so gatherthread is actively maintained.

GatherThread

Work together. Bring your Agents. Keep going anywhere.

GatherThread puts your team's discussion, AI-assisted work, and file review in one shared conversation. Collaborate live, use your own Agent, and continue from a computer, phone, or tablet. Less copying between chats. More progress everyone can follow.

Open GatherThread · Try the example · Illustrated guide · 简体中文

GatherThread: work together across devices, with your own Agents

Open source · Self-hostable · Current version identifier 0.1.0-beta.2 · Apache 2.0

Public-registration Beta: verify your email and choose a password to create an account. Available cloud models and allowances are shown in the workspace. Project invitations join a team; they are not account registration codes. Release history · Developer test environment

Your team. Your devices. Your Agents.

Work together in the same conversation. Agree on a goal, ask your Agent to work, and let teammates follow the result. Quote a particular message or mention someone instead of explaining the same thing twice. Discussion and development stay together.

Keep going beyond your desk. The same account can be signed in on computers, phones, and tablets at once. On your phone, continue the discussion, review progress, or ask your connected computer Agent to keep working. Move to another computer by downloading the file version you saved.

Choose the helper that fits. Connect Codex or DeepSeek Harness, or use the cloud trial Agent for a small task without local setup. Choose from that Agent's supported models and options. Each teammate keeps their own setup while sharing discussion and results; cloud usage allowances are shown in the workspace.

People share discussion and file versions while keeping their own Agents

A phone or tablet can request your computer Agent from another network. Keep that computer and its connector online, and use a GatherThread server the mobile device can reach. You do not install a local Agent on the phone.

One project, from discussion to delivery

These are full application screenshots with key controls annotated, from the disposable example. People, messages, and Agent replies are demonstration content.

1. Agree on what to make

Maya wants a club signup page. Alex confirms the time, place, and button behavior. Both see the same discussion. Sending Chat talks to people; it does not start an Agent.

Full interface with annotations: shared conversation and members

2. Turn decisions into work

Choose or connect your Agent, then explicitly ask it to build. The example shows a work plan, actual index.html and src/signup.js files, and behavior checks. Open Result to try the button or inspect the files. If discussion is long, summarize selected requirements with your local or cloud Agent; undeleted originals remain available.

Full interface with annotations: Alex's Agent answer and model

3. Check the changes before combining them

Each member uploads to their own branch. In the GT Cloud example, Maya submits a friendlier confirmation message for Alex to review. On GitHub, use the repository's pull-request review. Automatic upload does not mean automatic merge.

Full interface with annotations: reviewing Maya's file change

4. Continue on another device

Open the same conversation on your phone or tablet. Catch up, discuss the next step, or request your own online computer Agent without returning to your desk.

<img src="docs/assets/product/en/annotated/12-phone.jpg" width="390" alt="Full interface with annotations: the shared conversation on a phone">

Follow the complete story, including the working signup page, in the illustrated product guide.

Project files: GitHub for real development

Choose GitHub for ongoing development. Each member authorizes their local connection and uploads to a personal branch. For this local sync path, files go directly from the computer to GitHub and do not count against GatherThread cloud file quotas. Automatic upload waits for Agent work to stop and files to settle; manual upload, download, and recovery are also available. Review and integrate through GitHub.

Choose GT Cloud for a lightweight trial. Try a small project with GatherThread's file service: 128 MiB of current file versions per user, without first setting up GitHub. Inspect member versions, submit changes for review, let the project owner integrate them, and recover uploaded files.

The destinations are independent: connecting one does not move the other's files. File sharing needs separate authorization, and its upload switch is separate from conversation uploads. Neither path automatically downloads or merges. Use separate local folders for people or tools editing at the same time; personal branches do not isolate a shared local folder.

GitHub's permissions, platform rules, and local transfer safeguards still apply. Leaving a GatherThread project does not revoke GitHub access. These features save project source files, not a complete backup of your computer. See file setup, limits, and recovery.

Get started

  1. Sign in with a verified email address and password. New here? Verify your email, choose a password, then create a project and a Multi conversation. Invite your teammates, or accept an existing project's invitation.
  2. Choose your Agent. Try the cloud Agent without installing anything, or connect your computer with the Codex launcher or DeepSeek Harness plugin. A phone can use the cloud Agent directly, or that same account's connected computer Agent while the computer and connector stay online.
  3. Talk, then ask for work. Use Chat for discussion and the separate Agent request button for AI work. Select an available Agent and its supported model options before requesting it.
  4. Share files when you need to. Prefer GitHub for development; choose GT Cloud for a small trial. Authorize file access separately.

Want to practice first? Open the free example or the visual guides in Settings. Example changes are disposable, use no model quota, and do not affect real projects.

More ways to work together

| Capability | What you can do | |---|---| | Multi and Solo | Collaborate in Multi; write your own Solo while project members can still read it. Solo is not a private conversation. | | Quotes and mentions | Reply to a specific chat or Agent answer; open the mentions inbox to find messages addressed to you. | | Shared summaries | Ask your own local or cloud Agent to summarize selected messages, keep undeleted originals and valid older versions, and summarize again. | | Delete your own messages | Use the trash icon on your chat, Agent request or Agent reply. Related generated work and summaries may also become unavailable; other members' messages are not yours to delete. | | Conversation synchronization | Send shared history to connected Agents; switch local-to-cloud automatic uploads on or off per conversation, or manually upload completed turns. | | Codex history import | Create a new visible local task from shared history. Review it and archive the old task yourself; live history delivery remains independent. | | Models and request controls | Choose supported models and reasoning levels. During local or cloud Agent work, the request button becomes Pause or Resume. | | Cloud trial Agent | Try small tasks in an isolated workspace without local installation or your own API key; workspace usage allowances apply. | | Personal model APIs | Add your own model account in Settings for cloud tasks and summaries. Your provider bills you; keys are not shared with teammates. | | GitHub cloud workspace | With a validated cloud coding model and your repository authorization, request an npm project change, review its saved source and explicitly create a draft PR. | | Members and invitations | Manage owner, participant, and viewer roles; leave another person's project or remove members after resolving shared files. | | Accounts and devices | Sign in on multiple devices, revoke a device independently, change your profile, or review the impact before deleting your account. | | Bilingual onboarding | Use Chinese or English and learn through visual guides tailored to computers, phones, and tablets. | | Live work progress | Follow growing answers and expand the Agent's available, shareable thinking, tool work and plans. Collapse sections independently; missing information is never invented. | | Rich answers | Read Markdown, code and formulas without leaving the conversation. | | Self-hosting | Run the collaboration service on your own server. |

Summaries may omit details and use your Agent's quota. Later Agent requests default to the summarized history; Settings can select original history instead. Reading-view toggles are separate. Pause/Resume controls a GatherThread request, not an identical freeze and restart of a native Agent's internal reasoning.

Choose what you share. Shared discussion goes to GatherThread; files you authorize go to the selected file service. Sharing a conversation does not give teammates your local tool approvals or model credentials. Check files for secrets before uploading. Cloud deletion does not delete local Agent conversations or files. See privacy and security.

After message deletion, all future Agent history prepared by GatherThread will exclude the erased content and invalidated results. Deletion does not roll back file changes, erase other members' independent messages or manual reposts, or recall copies already held locally or by model providers. The live conversation retains a placeholder with its order, author and time; privacy explains retention.

Next: Agents working more directly together

Our direction is simple: let one person coordinate several Agents, let teammates' Agents collaborate directly, and let multiple cloud Agents work as a team. These automatic team capabilities are future plans, separate from today's shared conversations and individually requested Agents.

Explore, contribute, or self-host

Product guide · All documentation · Codex setup · DSH setup · File collaboration · Self-hosting · Release notes

Developers: start with AGENTS.md and CONTRIBUTING.md, then the architecture and interface contracts. Use Node.js 24 or newer; see the hosting guide for local setup. The project lead reviews version updates before release.

Report a bug or suggest an improvement. Report security issues privately to coolhezi@sjtu.edu.cn; never post credentials or private transcripts.

Licensed under Apache 2.0. Copyright 2026 Yuhan He and contributors.

Related Skills

View on GitHub
GitHub Stars10
CategoryDevelopment
Updated1d ago
Forks1

Languages

TypeScript

Trust signals

97/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 info