← All docs
AgentProfile

Portable PROFILE.md files for AI agents.

AgentProfile is a lightweight PROFILE.md format and Python parser for defining an agent's identity, model, tools, skills, provider options, output schema, and system prompt in one version-controlled file.

Create a PROFILE.md

The primary artifact is the file. The Python package is optional tooling for validation, parsing, and serialization.

PROFILE.md example

---
name: researcher
description: Research agent
model: openai:gpt-4o
tools:
  - web_search
skills:
  - literature-review
provider: provider.json
output_schema: output-schema.json
---

You are a careful research agent. Cite sources and hand off open questions.

Load a profile

Install from GitHub while the parser package is early:

pip install "agentprofile @ git+https://github.com/open-assistants-lab/AgentProfile.git"
from agentprofile import load_profile, dumps_profile

profile = load_profile("./researcher/PROFILE.md")
print(profile.name)
print(profile.system_prompt)

markdown = dumps_profile(profile)

Core concepts

  • PROFILE.md contains frontmatter metadata and a Markdown system prompt body.
  • provider.json can hold provider-specific model options.
  • output-schema.json can define a structured output contract.
  • The format is runtime-neutral; runtimes decide how to execute the profile.

Fields

  • name, description, model
  • tools, skills, tags
  • handoff_instructions
  • provider, output_schema

Status

AgentProfile is an early prototype. The schema is subject to change as we learn what fields matter for multi-agent team formation. The format is not yet stable and should not be treated as a standard.