Instructions
DAL gives agents clear instructions so they can perform tasks step by step.
How DAL turns a goal into durable work
Student
Writes the goal
DAL program
Loads context
One handoff
A single next step
Evidence
Files you can reopen
The basics
What is Distributed Agent Language?
Tools
Agents can use tools to read files, run commands, search, and create outputs.
Collaboration
Multiple agents can share work, divide responsibilities, and finish bigger projects.
The contract
Five steps from idea to artifact
This is the DAL habit. Chat can help you think. DAL is for work you can inspect, resume, and hand to someone else.
-
1. Write the goal.
Put it in a file such as
GOAL.md. If it only lives in chat, the next session has to guess. - 2. Load context from files. Read what you already have — notes, a deadline, a syllabus — instead of pasting everything into a prompt.
- 3. Make one bounded handoff. One output, one owner, one next step. Do not start a second plan in the same run.
- 4. Leave evidence. Write a log or receipt so you can see what loaded, what was written, and what to do next.
-
5. Degrade instead of crashing.
If a file is missing, finish with an explicit
partialoutcome. Keep what did load.
Do this
Try the study-plan sample
A student-week planner: load notes and a deadline, write exactly one study-plan handoff, and log the outcome. Toggle the deadline off to see a partial run.
handoff.md
work_log.md
Run it on your machine
Install DAL, then from the students/sample folder:
dal run sample.dal
Rename or remove deadline.md and run again to see "outcome": "partial". Open the files in DAL IDE if you prefer a browser editor.
Why start now
First-mover advantages for students
Stand out early
Students who learn agent workflows early can build stronger portfolios and practical demos before the field becomes crowded.
Learn transferable skills
DAL practice reinforces planning, automation, debugging, and systems thinking—skills that apply across software, data, and AI roles.
Move from idea to prototype faster
Using a structured agent language reduces setup friction and helps teams turn concepts into working prototypes more quickly.
Getting started
A simple learning path
- 1. Read the idea. Understand what an agent is and how DAL describes actions. Start with What is DAL?
-
2. Run the sample.
Use the study-plan sample on this page, then
dal run sample.dalon your machine. - 3. Build something small. Reuse the same contract: one goal file, loaded context, one handoff, one log. A study helper or research summary is enough.
- 4. Improve it. Add another context file, a missing-file path, or a second agent that only receives the handoff — do not skip the evidence.
Why this is real
COO is the proof of concept — not the homework
The sample above is DAL: a language and runtime for explicit, file-backed agent work. COO is a local operator built on that same contract — queues, owners, review gates, and work that survives overnight. You do not need COO to learn DAL. It is the existence proof that the five-step habit already runs a real system.
The Operator bundle can still fit a student who wants that loop on a laptop: a queue for class projects, drafts that stay in a folder, and send that waits for you. You do not have to write DAL to use it — the language runtime is already in the kit — so the sample on this page remains the way to learn the language, and the bundle is optional if you later want an operator around the same kind of work.
if you want to help us test and improve by providing feedback and reporting issues.
Terms and questions
Glossary and FAQ
Glossary
- Agent
- A program that receives a task, reasons about it, and can take actions using tools.
- DAL
- Distributed Agent Language — a structured way to describe what agents should do and how they coordinate.
- Artifact
- A durable file the work leaves behind — a plan, draft, handoff, or log — that another person or agent can open later.
- Handoff
- One bounded output for a run: who it is for, what was done, and the single next step. Not a chat transcript.
- Tool
- A capability an agent can call, such as reading a file, searching the web, or running a command.
- Partial
- A finished run that names missing context instead of failing. Loaded files are kept.
- Workflow
- A sequence of steps — often across multiple agents — that completes a larger task.
FAQ
Do I need to be an expert programmer to start?
No. Start with the sample on this page. You can read the files first, then install DAL when you want to run sample.dal yourself.
How is DAL different from a regular chatbot?
Chatbots mainly generate text. DAL programs load files, write artifacts, and leave a trail so the next session does not start from zero.
What should my first project be?
Run the study-plan sample, then copy the same five steps onto one assignment you already have: one GOAL.md, your notes, one handoff.md.
Do I need COO to use DAL?
No. DAL is the language. COO is a proof-of-concept operator that uses DAL. Learn the sample first; look at COO later if you want a full local operator.