Introduction to Eve
Eve is Vercel's open-source, filesystem-first framework for building, deploying, and scaling AI agents with production infrastructure built in.
$ open "Run your first eve agent"
Meet eve, Vercel's filesystem-first agent framework. Launch a real eve session in the terminal and have your first conversation with the...
checkpoints run as you work
✓ checkpoint passed, lab 2 unlocked
What you'll build
Curriculum
12 labs, about 2 h
- 1
Run your first eve agent
Meet eve, Vercel's filesystem-first agent framework. Launch a real eve session in the terminal and have your first conversation with the agent.
6 min
Pro - 2
Watch eve use its tools
Eve agents act through tools in an isolated sandbox. Ask the agent to create a file and watch it call write_file, then see the result appear in the workspace.
8 min
Pro - 3
Have eve build something real
Describe an outcome and let the agent produce it. Ask eve to build a small HTML page and watch the finished file appear in your workspace.
9 min
Pro - 4
Give your agent a custom tool
Tools are how eve agents do real work. Drive the agent to call a custom save_note tool and watch a real note file appear in the project.
9 min
Pro - 5
Teach your agent a skill
Skills are load-on-demand know-how. Trigger a filing skill and watch the agent follow a procedure it pulled into context, footer and all.
9 min
Pro - 6
Delegate work with subagents
Eve agents delegate focused work to subagents that run with fresh context. Have the agent fan out three notes in parallel and gather the results.
10 min
Pro - 7
Add a human in the loop
Gate risky actions behind human approval. Ask the agent to publish a note, watch the run pause, approve it, and see the note go live.
10 min
Pro - 8
One session, real memory
An eve session is durable and remembers context across turns. Tell the agent a fact, then have it use that fact in a later message to write a file.
8 min
Pro - 9
Prove it works with evals
Evals are scored checks that catch regressions when you change an agent. Create and run a real eval against the live agent.
11 min
Pro - 10
Connect eve to the world
A connection is one file that points your agent at an external MCP server or API. Wire up a local MCP server and watch your agent gain a new tool with no restart.
11 min
Pro - 11
Put your agent on a schedule
Schedules let an eve agent wake on a cron and work on its own, and one deploy ships it all. Create a real scheduled job file.
9 min
Pro - 12
Your eve capstone
Combine a custom tool, a scheduled job, and an eval in one request so the agent leaves behind production-shaped artifacts.
12 min
Pro
Your instructor
Elvis Saravia
Founder, DAIR.AI
Elvis founded DAIR.AI and wrote the Prompt Engineering Guide, one of the most widely used references on working with language models. He has worked at companies like Meta AI and Elastic, and specializes in AI agents, RAG, and harness engineering.
How labs work
Live workspace
You run the real tools in a browser workspace. Nothing to install.
Checkpoints
Each lab checks your files and output, then unlocks the next one.
Your work stays
The workspace persists between labs so you finish with a project.