Amp News (Sourcegraph)→ original

Sourcegraph's Amp Adds Global Plugins and Skills for AI Agents

Sourcegraph's AI coding agent Amp now supports global plugins and skills. Users can store them in the cloud, import them from repositories, share them with their team and update them with a single command. There are two tiers: personal plugins for experimentation and team plugins for the whole organization. Plugins can be remixed and published back to the workspace.

AI-processed from Amp News (Sourcegraph); edited by Hamidun News
Sourcegraph's Amp Adds Global Plugins and Skills for AI Agents
Source: Amp News (Sourcegraph). Collage: Hamidun News.
◐ Listen to article

In August 2026, Sourcegraph added global plugins and skills to its AI coding agent Amp. Extensions are now stored in the company's cloud and work across all projects and devices, with no manual syncing of configuration files.

How do global plugins and skills work in Amp?

Global plugins and skills are cloud-based extensions for the Amp agent. They work wherever Amp runs: on a local machine, in CI and in orbs environments. Previously, users kept plugins locally in the `~/.config` directory. That made them unavailable to other team members and required manual syncing whenever someone switched devices.

In Amp's terminology, a plugin is a set of instructions and tools that change the agent's behavior. A skill is a more specific capability: running a formatter on every file the agent edits, testing a browser, or generating release notes from a template. Through the agent, users can now do the following in natural language:

  • create a personal plugin that runs a formatter on every file the agent edits;
  • import a browser-testing skill from a repository into their personal skills for all projects;
  • find out whether anyone on the team already has a skill for writing release notes;
  • check whether imported plugins are out of date.

Workspace administrators get an additional option: copying a specific member's plugin into the team workspace. Management happens through the User Settings and Workspace Settings sections of the Amp interface.

How do personal plugins differ from team plugins?

Personal plugins are meant for experimentation. A user can ask the agent to build something new and immediately reload plugins in the same thread, without restarting the development environment. That shortens the iteration cycle to a few seconds.

Team plugins are published by workspace administrators and load by default for the entire team. Sourcegraph recommends a clear sequence: test a plugin in your personal space first, confirm it works correctly, and only then publish it to the team. If an existing team plugin needs major changes, import it into your personal space, rework it and publish the updated version.

"Personal plugins and skills are much more than a replacement for `~/.config`: you can share them with your team so people can discover, import and remix them," according to the official

Amp blog.

Remixing and updating with a single command

Amp supports full sharing and remixing of plugins. A user can open a personal plugin to the workspace, and other members can find it, import it and adapt it to their own tasks. If the original receives updates, running `amp skill update <name>` or `amp plugins update <name>` is enough. Users can also ask the agent to update a skill to the latest version.

According to the Amp blog, global plugins are becoming more than a replacement for local configuration files. They are turning into a full platform for discovering and exchanging extensions inside an organization, similar to npm packages or VS Code extensions.

What this means

Agentic development environments are growing extension ecosystems, much as IDEs once did. Team-wide sharing and remixing of plugins lowers the barrier to entry for everyone: there is no need to write configuration from scratch, since a colleague's ready-made skill can simply be imported and adapted. This is especially valuable for companies with dozens of developers, where standardizing agent behavior pays off organizationally. If this pattern spreads to other AI development tools, "installing a plugin for the agent" could become as routine within a couple of years as adding an extension in VS Code.

ZK
Hamidun News
AI news without noise. Daily editorial selection from 50+ sources. A product by Zhemal Khamidun, Head of AI at Alpina Digital.
✦ Autonomous AI Agents

Delegate Deep Research, Coding & Workflows to an AI Agent

Deploy autonomous multi-step execution with NomadNet Agent (agent.nomadnet.ai) — automated web research, data synthesis, code generation, and complex task orchestration.

Launch NomadNet Agent →NomadNet AI Platform
What do you think?
Loading comments…