Dana: The World’s First Agentic OS
Build deterministic expert agent easily with Dana.
A complete Expert Agent Development Toolkit: Agentic out of the box. Grounded in domain expertise.
Why Dana?
Most frameworks make you choose:
- Too rigid → narrow, specialized agents.
- Too generic → LLM wrappers that fail in production.
- Too much glue → orchestration code everywhere.
Dana gives you the missing foundation:
- Deterministic → flexible on input, consistent on output — reliable results every run.
- Contextual → built-in memory and knowledge grounding let agents recall, adapt, and reason with domain expertise.
- Concurrent by default → non-blocking execution; agents run tasks in parallel without threads or async code.
- Composable workflows → chain simple steps into complex, reproducible processes that capture expert know-how.
- Local → runs on your laptop or secure environments, ensuring privacy, speed, and mission-critical deployment.
- Robust → fault-tolerant by design, agents recover gracefully from errors and edge cases.
- Adaptive → agents learn from feedback and evolving conditions, improving performance over time.
Install and Launch Dana
💡 Tip: Always activate your virtual environment before running or installing anything for Dana.
# Activate your virtual environment (recommended)
source venv/bin/activate # On macOS/Linux
# or
venv\Scripts\activate # On Windows
pip install dana
dana studio # Launch Dana Agent Studio
dana repl # Launch Dana Repl
- For detailed setup (Python versions, OS quirks, IDE integration), see Tech Setup.
What’s Included in v0.5
Agent Studio
Turn a problem statement into a draft expert agent with three parts — agent, resources, workflows. Studio generates a best-match workflow and lets you extend it with resources (documents, generated knowledge, web search) or edit workflows directly.
Agent-Native Programming Language
A Python-like .na language with a built-in runtime that provides agentic behaviors out of the box — concurrency, knowledge grounding, and deterministic execution — so you don’t have to wire these up yourself.
What this means for you: You can build and iterate on expert agents faster, with less setup and more confidence they’ll run reliably in production.
Full release notes → v0.5 Release.
First Expert Agent in 4 Steps
-
Define an Agent
agent RiskAdvisor
-
Add Resources
resource_financial_docs = get_resources("rag", sources=["10-K.pdf", "Q2.xlsx"]) -
Follow an Expert Workflow
def analyze(...): return ... def score(...): return ... def recommend(...): return ... def wf_risk_check(resources) = analyze | score | recommend result = RiskAdvisor.solve("Identify liquidity risks", resources=[resource_financial_docs], workflows=[wf_risk_check]) print(result) -
Run or Deploy
dana run my_agent.na # Run locally dana deploy my_agent.na # Deploy as REST API
Learn More
- Core Concepts → Agents, Resources, Workflows, Studio.
- Reference → Language syntax and semantics.
- Primers → Deep dives into Dana language design.
Community
Enterprise support
License
Dana is released under the MIT License.
© 2025 Aitomatic, Inc.
2 maintainers
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file dana-0.6.0.3.post3.tar.gz.
File metadata
- Download URL: dana-0.6.0.3.post3.tar.gz
- Upload date:
- Size: 10.8 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.2
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
e817f2c55d587bf577fc56cd6408d57b89d85f8e4ff266c960d3f241f9ccca24
|
|
| MD5 |
d199c8204b7a6e746531ff1631ec244f
|
|
| BLAKE2b-256 |
6998e73653f6d2743ea4293c0043cb7a49989c16ab43b27972c720031cecf285
|
File details
Details for the file dana-0.6.0.3.post3-py3-none-any.whl.
File metadata
- Download URL: dana-0.6.0.3.post3-py3-none-any.whl
- Upload date:
- Size: 11.7 MB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.2
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f5ca3589bb81769abfb8033a3246a14bcc8406eb85a407ee39308d1ef4dfedd3
|
|
| MD5 |
382debfe99fed01c183a1f9fae9e7abd
|
|
| BLAKE2b-256 |
f7b08260d71546e7b80fa995512d6c7cacb4aa3b68cee7e650a29d973fd05154
|
