.png)
4 months ago
22
In early 2024, we witnessed some of the most significant technical breakthroughs since the advent of the internet. Artificial Intelligence arrived in full force… and it’s clear that it isn’t going anywhere. While there’s still some uncertainty about the full extent of what AI can do, the initial fear of robots taking over seems to have passed.
For me, the spark hit in early 2023. I caught the AI bug hard. At first, I was obsessed with replacing my Google searches with ChatGPT prompts. That quickly snowballed into diving deep into model comparisons (which ones were fastest, most accurate, most powerful)… you know the drill. But then, I discovered vibe coding.
When I found Cursor in 2024, everything changed. Suddenly, my lack of deep technical skills wasn’t a blocker anymore. Like many folks with ADHD working in tech, I have a notes app overflowing with half-baked app ideas and little tools to solve everyday problems. But for the first time, I realized I could actually build them on my own.
So I carved out time between work and family life to build micro-projects whenever I could. I fell in love with the technology, the community, and most importantly, the energy. I was all in.
Fast forward to May 2025: Sourcegraph launches the research preview of Amp, an in-editor AI coding agent. Could it get any better? Actually, yes. Back in late 2024, I joined Sourcegraph (spoiler alert) and today, I have the privilege to be part of the team that is bringing this groundbreaking product to market!
That’s why I’m excited to share My Top 5 Favorite Features of Amp. The features that, in my opinion, make it a must-have in any modern developer’s toolkit.
The future of AI coding isn’t coming. It’s already here. It’s Amp.
Before I dive into my favorite features and how I use them in my day-to-day life, I want to highlight a few core features and foundational categories that set Amp apart.
First, Amp is architected to be IDE agnostic. Unlike Cursor or Windsurf, which are built on a fork of VSCode, Amp isn’t tied to any single development environment. Meaning, if you want to connect to an MCP server you don’t have to wait for features to ship. Amp is available both as a VSCode extension and a CLI tool, giving developers true flexibility to use it wherever they work best. You’re not locked into a specific editor — you can integrate Amp seamlessly into your existing workflows.
Second, Amp is outcomes obsessed. It’s not about locking into a single model provider — it uses the best model for each specific task. There are no artificial token constraints either; Amp prioritizes delivering high-quality code results, even if that means allocating more compute. The focus is on output quality, not cost-saving compromises.
Third, Amp was built for teams from day one. It goes beyond individual productivity and introduces features like thread sharing for collaborative development, team leaderboards to track engagement, and native support for team workflows. Where many tools are designed for solo developers and only later add team features, Amp was engineered with enterprise collaboration in mind from the start.
Fourth, Amp was architected with agents at its core. It wasn’t retrofitted from an autocomplete or copilot tool — it was built from the beginning as an autonomous agent. That design choice enables it to handle complex, multi-step tasks and allows its capabilities to evolve continuously as agent tech improves.
Finally, Amp is built on an enterprise-grade foundation. Developed by Sourcegraph — a trusted leader in code intelligence — it leverages advanced codebase context technology to deliver secure, scalable, and compliance-ready experiences for large teams. Whether you’re part of a startup or a Fortune 500, Amp is built to meet enterprise demands without compromising flexibility or power.
These foundational pillars are what make Amp stand out — not just as a tool, but as an extensible platform built for modern development.
When I say “my” top 5, that’s a bit misleading. These aren’t just the features I personally use the most, they’re the features our customers call out as game changers. After seeing how teams adopt Amp in the wild, it’s clear which features actually move the needle, not just in demos but in day-to-day workflows.
To set the groundwork: Amp can, and will, improve developer productivity from day one. Like any new tool, there’s a slight learning curve, but it’s minimal compared to the results your team will see. Whether you’re an individual dev with a growing side project or part of a large engineering org, the impact is immediate. These are the five features I believe make that possible.
Explanation of this feature: Unfettered token usage means no forced prompt fragmentation
The power: Handle entire projects in one coherent conversation
My biggest gripe with tools like Cursor, Windsurf, ChatGPT, and pretty much every other solution on the market today is token constraints. There’s nothing more frustrating than crafting a beautifully complex, detailed prompt… only to have it shut down because the thread has hit its limit.
Now you’re forced to start a new thread, which no longer contains the context you’ve built up. It’s disruptive, it’s tedious, and honestly, it defeats the whole purpose of using an AI assistant in the first place.
Look, if I’m only paying $20, $30, or even $50 a month, then sure — there are going to be limits on how many credits I can burn through. The financial trade-offs are inevitable.
But with Amp, we said: to hell with that.
“Every year we have a substantial budget for outsourcing dev work to contractors. This requires us to train them on our branding and onboard them across different projects… Amp can do all of this work in less time with much less of a commitment”
Anu Choudhury @ Reputation
An example prompt:
The result w/ summary:
Amp vs Competition (workflow):
Amp approach:
- “Analyze entire platform architecture: auth, payments, notifications, reporting”
- Single comprehensive analysis
- Full system understanding maintained
- No artificial workflow interruptions
Others approach:
- “Analyze auth system” → Hit token limit → New thread
- “Now analyze payments” → Hit limit again → Another thread
- “How do they connect?” → Lost context from previous threads
Explanation of this feature: Multiple threads running simultaneously for complex, multi-workstream projects
The power: 10x productivity through parallel agent execution
“I fired up 4 different threads with 4 distinct but inter-related tasks hit cmnd-enter on all four and 10 minutes later everything was done… being able to think AND execute in parallel felt like I finally had an extension of my brain and not just another tool to use…blew my damned mind”
- Connor O’Brien @ Sourcegraph
Real enterprise scenario:
- Thread 1: “Refactor entire user authentication system — migrate from JWT to OAuth2”
- Thread 2: “Update all ‘Submit’ buttons to new design system — 47 components across webapp”
- Thread 3: “Migrate legacy Python 2.7 scripts to Python 3.11–23 files in data pipeline”
Hit Command+Enter on all three threads
→ All agents start working simultaneously
→ 10 minutes later: Complete system transformation
Explanation of this feature: Model Context Protocol support for seamless tool integration
The power: Plug-and-play extensibility without complex setup or API wrangling
Amp’s Advantage over Competitors:
Amp approach:
- Add any MCP-compatible tool instantly
- Internal tools, external services, custom workflows
- Infinite extensibility through open protocol
Competitors:
- Request new tool or feature from vendor
- Wait for official integration
- Limited to pre-approved tools
- No access to internal company tools
Enterprise use cases:
- Security teams add vulnerability scanning MCP servers
- DevOps teams integrate CI/CD pipeline controls
- QA teams connect test automation frameworks
- Compliance teams add audit trail and approval workflows
- Architecture teams integrate design decision databases
Explanation of this feature: Conversations that become searchable institutional memory
Unlike Cursor’s ephemeral chats, Amp threads are permanent team assets.
Real workflow:
- Senior Engineer: “Why did we choose PostgreSQL over MongoDB for user sessions?”
- Searches team threads: “database selection user sessions”
- Finds 3-month-old thread with complete analysis
- Decision rationale, benchmarks, and trade-offs all preserved
- No need to re-research or guess at past decisions
Enterprise value:
- Knowledge retention when team members leave
- Decision audit trails for compliance
- Onboarding acceleration — new hires search existing threads
- Reduced duplicate work — solutions already documented
Explanation of this feature: One-click context sharing that transforms how teams collaborate
The power: Share entire investigation context, not just final code
Traditional code review:
- PR: “Refactored auth service”
- Reviewer: “Why this approach?”
- Author: “Let me explain…” (15-minute Slack thread)
Amp thread sharing:
- PR: “Refactored auth service — full analysis: https://ampcode.com/thread/abc123"
- Reviewer: Clicks link → views completed investigation
Result: Instant context, faster approval, better decisions
Game-changing scenarios:
- Architecture decisions — share the complete thought process
- Bug investigations — show the entire debugging journey
- Performance optimizations — document the full analysis
- Security reviews — provide complete threat modeling context
- Team Leaderboards: Gamifying Excellence
- What it is: Real-time tracking of thread activity and knowledge contributions
- The power: Incentivizes documentation and highlights subject matter experts
What it is: Built-in task tracking and progress management within threads
The power: Seamless project coordination without external tools
Workflow integration:
“Analyze entire platform architecture: auth, payments, notifications, reporting”
→ Amp automatically creates TODOs for each service
→ Tracks progress: “Auth Service A: Completed”, “Auth Service B: In Progress , etc..”
→ Team visibility into complex task breakdown
→ No external project management tools needed
I know, I said top 5… too many great features to stick to just 5.
What it is: Complete tracking and reversion of agent-made changes
The power: Enterprise compliance and change management
Enterprise workflow:
Agent modifies x number of files during refactoring
→ Hover shows exact changes per file
→ Revert individual files or entire changeset
→ Edit message automatically reverts subsequent changes
→ Complete audit trail for compliance
I hope you enjoyed this post. This evolution in how we build products and do business is just getting started, and I can’t wait to see what other teams create with Amp.
If you liked what you read and want to hear more of my thoughts on the evolution of this space, follow me on Twitter (are we still calling it that?) or connect with me on LinkedIn. And if you’re ready to try Amp for yourself, you can get started for free at ampcode.com.
