Our take
Newless than six months old and changing fast- Activity
- 68
- Community
- 4
- Issues
- 95
- Pull requests
- 86
+0 stars in the last 7 days
- Merges more pull requests than 100% of the projects we track
- Closes issues faster than 82% of the projects we track
- Fewer contributors than 92% of the projects we track
This is a broad, actively changing agent framework with useful runnable compositions, but its newness, bug-fix-only policy, and dependence on one contributor make adoption a deliberate choice.
Good fit if
- You want agent tools and state transitions to be profile configuration interpreted by a Go runtime.
- Your team can evaluate a fast-changing project with recent releases and rapid issue and pull-request handling.
- BSD-3-Clause licensing fits your distribution and compliance needs.
Look elsewhere if
- You need a project that is accepting feature development, since the repository says it accepts bug fixes only.
- You require a mature contributor base; current project activity is concentrated in one contributor.
- Your agent design does not benefit from YAML-defined workflows or the included Go runtime and application model.
All health signals
| Last commit | 2026-09-24 (16 days ago) |
|---|---|
| Commits, last 90 days | 100+ |
| Releases, last 12 months | 2 (latest ui-kit/v1.1.0, 2026-09-19) |
| Contributors | 1 |
| Issues closed, last 90 days | 76+ (typically closed in 0 days) |
| Pull requests merged, last 90 days | 100+ (typically merged in 0 days) |
| Project age | 3 months |
Checked on 2026-10-11 with the GitHub API.
Overview
Declarative Agents is a Go runtime for tool-augmented LLM agents whose workflows are described in YAML profiles rather than fixed in application code. Profiles specify tools, states, transitions, signals, and budgets, so a workflow change can be made as configuration instead of requiring a new runtime binary.
The repository includes reusable agent and tool blocks, runnable applications, and a white paper describing design patterns for dependable agents. It is aimed at developers building configurable agent systems who are prepared to work with the repository's profile model and Go runtime.
Key Features
- Defines agent behavior in YAML profiles interpreted by one Go runtime.
- Loads profile, machine, and tool definitions, with a
--validate-configmode for checking configuration without serving. - Provides a standard tool library and reusable agent and tool blocks in a catalog.
- Includes example applications, including a planner-executor-critic coding agent and a documentation-curator presentation.
- Offers build, audit, test, and integration commands through Mage across repository modules.
- Documents eleven design patterns and provides a normative application DSL specification.
Use Cases
- Agent-platform developers can change state-machine workflows through profiles rather than rebuilding runtime code for every loop change.
- Teams composing agents from reusable tools can use the catalog to share blocks across applications.
- Developers evaluating multi-step coding-agent workflows can examine the planner-executor-critic application and its kind integration gates.
- Teams exploring declarative agent design can use the white paper and DSL specification alongside the runnable examples.
What you need
Detected in the repository
- Go (main language on GitHub)
License in plain words
BSD-3-Clausepermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice; the authors' names cannot endorse your product
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
The README's quickest runnable example is:
cd applications/agent-architecture
mage run
This requires Go and Mage. See the repository README for direct profile validation, other applications, and module-specific commands.
Alternatives
- mcp-agent: mcp-agent is a Python framework centered on MCP server integrations and composable workflows, rather than YAML-defined agent state machines in Go.
- autogen: AutoGen is a Python framework for conversational multi-agent collaboration, rather than a Go runtime for YAML-defined profiles.
- openai-agents-js: OpenAI Agents JS is a TypeScript SDK focused on tool use, delegation, and conversations, rather than YAML-configured state machines.
- griptape: Griptape is a modular Python framework for tasks, pipelines, and workflows, rather than a Go runtime that interprets YAML agent profiles.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| declarative-agents | Go | BSD-3-Clause | 4 | New |
| mcp-agent | Python | Apache-2.0 | 8.6k | Quiet |
| autogen | Python | CC-BY-4.0 | 61.3k | Quiet |
| openai-agents-js | TypeScript | MIT | 3.9k | Active |
| griptape | Python | Apache-2.0 | 2.6k | Active |
Considerations
The repository is new and changing quickly, with frequent recent commits and releases, but that activity should not be mistaken for a long track record. The project reports one contributor, with work concentrated in that person, which creates continuity risk. Its maintenance-mode notice also limits expected development to bug fixes. The README describes extensive audit and integration gates, but the measured requirements data does not report tests or CI as project requirements; verify the gates relevant to your intended application before relying on them. The BSD-3-Clause license permits broad use subject to its notice and disclaimer conditions.
Found this useful?
Share it with someone who would like declarative-agents.