Your First Agent¶
Welcome to Dokugent! This guide walks you through creating and certifying your first AI agent using Dokugent CLI. You’ll define its behavior, evaluation standards, and generate a certifiable contract.
1. Initialize the Workspace¶
Creates the .dokugent/ folder and assigns a timestamped agentId (birthstamp).
2. Register the Agent¶
Provide the agent name, purpose, and tags.
3. Link an Owner¶
Select or create the identity used for signing certificates.
4. Generate a Signing Key¶
This creates a public/private keypair used to sign and verify your agent certificates. You only need to do this once per identity.
5. Plan Your Agent¶
Define your agent’s task steps (e.g., summarize_input, web_search).
6. Define Output Criteria¶
What does a “good” output look like? Choose quality benchmarks like clarity, accuracy, safety.
7. Add Conventions¶
Model-specific behavior guides (e.g., dev, research, writing).
8. Optional: Fill Out I/O and Compliance¶
Before previewing or certifying, we recommend filling out:
.dokugent/data/io/– rules about expected file inputs/outputs, file type restrictions, etc..dokugent/data/compliance/– metadata like contact info, legal basis, user roles
These aren’t blockers, but they’re important for agents that run in structured or regulated environments. Think of them like your agent’s operating policy, not unlike a terms-of-use document.
9. Preview the Agent¶
Simulate your agent’s behavior and verify that criteria and plan steps align.
10. Certify It¶
This signs a .cert.json file, verifying the plan, criteria, owner, and birthstamp.
11. Compile the Bundle¶
Outputs a final .compiled.cert.json for runtime usage, including any BYO bundles, hashes, and logs.
You’re now ready to expand your agent with tool access, security checks, and multi-step simulations!