AgentDock
Build AI agents with configurable determinism, balancing creativity and reliability.
- Source repo
- AgentDock/AgentDock
- Stars
- ★ 1.8k
- Last updated
- 3mo ago
- License
- MIT
- Primary language
- MDX
- FA score
- 32/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformOpenAI API · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Developers using TypeScript to build custom AI agents who want to balance predictability with creativity through configurable determinism and multi-step tool calls.
- Not a fit if
- Teams needing stable APIs: Beta status, core package not yet a stable NPM release
- Users wanting out-of-the-box setup: multiple LLM and tool API keys configured manually
- Projects needing vector storage or long-term memory: still in development
- Source review
- 32/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
AgentDock is an open-source framework for building sophisticated AI agents with an emphasis on configurable determinism—balancing the creativity of LLMs with predictable system behavior. It consists of two main components: AgentDock Core, a backend-first framework for building and deploying AI agents, and an Open Source Client, a complete Next.js reference implementation available at hub.agentdock.ai. The framework is built around a node-based architecture, with core components like BaseNode and AgentNode, and supports tools as specialized nodes. It offers configurable deterministic workflows, multi-step tool calls, session management, an orchestration framework, and an evaluation framework. It is framework-agnostic, supports multiple LLM providers, and is released under the MIT License.
AgentDock provides composable nodes (e.g., AgentNode, BaseNode) to build AI agents, managed through node and tool registries. AgentNode orchestrates LLM interactions, tool usage, and agent logic, supporting multi-step tool calls. The framework supports deterministic workflows (defined by connected node execution paths) and non-deterministic agent behavior (driven by LLM reasoning), allowing developers to combine both in the same system. It includes session management (isolated state for concurrent conversations), orchestration (dynamic control of tool availability based on context), and an evaluation framework (to measure and improve agent quality). The Open Source Client, a Next.js app, serves as a reference implementation consuming these capabilities, and supports multiple LLM providers (e.g., Anthropic, OpenAI, Gemini, DeepSeek, Groq) configured via environment variables or UI.
- Developers using TypeScript to build custom AI agents who want to balance predictability with creativity through configurable determinism and multi-step tool calls.
- Teams quickly prototyping AI-powered applications by starting from example agent templates (e.g., medical diagnosis, cognitive reasoning, history mentor, calorie vision).
- Architects looking to integrate AI into existing Node.js backends while controlling provider dependencies (BYOK) and leveraging session management and orchestration features.
- QA engineers who need an evaluation framework to systematically measure agent performance against expected behaviors.
- Developers building automation pipelines (e.g., data processing, document analysis) that combine deterministic workflows with LLM components.
How do you install or deploy this agent?
Clone the repository: git clone https://github.com/AgentDock/AgentDock.git
Enable and prepare pnpm: corepack enable && corepack prepare pnpm@latest --activate
Install dependencies: pnpm install
Copy the environment file: cp .env.example .env.localAdd LLM provider API keys (at least one, e.g., ANTHROPIC_API_KEY, OPENAI_API_KEY) and tool-specific keys (e.g., SERPER_API_KEY, FIRECRAWL_API_KEY).
How do you use this agent?
Start the development server: pnpm devConfigure API keys either in the UI or via environment variables (BYOK model).
Explore example agent templates in the agents/ directory and select or create one that fits your needs.
Refer to the documentation in the docs/ folder and at hub.agentdock.ai/docs for creating custom nodes and using the framework.
What are this agent's strengths and limitations?
- Configurable determinism: allows mixing non-deterministic LLM reasoning with deterministic workflows in the same system, enabling predictable outcomes.
- Framework-agnostic: supports multiple LLM providers (Anthropic, OpenAI, Gemini, etc.), reducing vendor lock-in.
- Modular node architecture: highly extensible through custom nodes and tools by extending BaseNode.
- Built-in advanced features: session management, orchestration, and evaluation framework support production-grade agents.
- Beta status: the project is in Beta, the core package is not yet published as a stable NPM package, and APIs may change.
- Initial setup complexity: requires Node.js ≥20.11.0 and pnpm, and manual configuration of multiple API keys.
- Documentation is scattered: distributed between the repository and external sites, which may hinder quick onboarding.
- Incomplete features: storage abstraction layer and vector integration are still in development, possibly limiting advanced use cases.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AgentDock This agent | 32 · Major gaps | FrameworkFree + model costs | ★ 1.8k | 3mo ago | MDX | OpenAI API · Claude API |
| Pi Web | 53 · Major gaps | CLIFree + model costs | ★ 7.5k | today | TypeScript | — |
| MiniCode | 50 · Major gaps | CLIFree + model costs | ★ 1.1k | 10d ago | TypeScript | Claude API |
| Open Multi-Agent Canvas | 43 · Major gaps | Web appFree + model costs | ★ 551 | 4mo ago | TypeScript | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README claims built-in security features (e.g., API key encryption) but provides no implementation details or code evidence. No specific implementation of least privilege, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution was found. Therefore all trust criteria score 0.
Evidence shows: README and package.json provide consistent installation and running instructions, but no actual runtime verification. Dependencies are listed in package.json but availability is not verified. Error handling is mentioned in README but no specific implementation is provided. Therefore self-consistency, dependency availability, and failure messages each score 1.
Evidence shows: README describes multiple use cases (e.g., AI applications, integrations, automation) and provides example agents. Capability boundaries are mentioned in documentation but not clearly defined. Trigger precision is not detailed. Environment fit is addressed with Node.js and pnpm requirements and deployment options. Therefore audience and scenarios score 2, capability boundaries and trigger precision each score 1, and environment fit scores 2.
Evidence shows: README provides clear information architecture with directory structure and documentation links. Installation notes are detailed. Naming stability is not explicitly stated. Examples and FAQ sections provide example agents. Known limitations are mentioned in roadmap. License is MIT. Versioning and changelog are not provided. Maintenance responsibility is not clear. Therefore information architecture, install notes, examples and FAQ, and license each score 2, while naming stability, known limitations, versioning and changelog, and maintenance responsibility each score 1.
Evidence shows: README describes output capabilities but provides no concrete examples. Marginal value is reflected in feature list. Cost-benefit is not discussed. Therefore output usability scores 1, marginal value scores 2, and cost-benefit scores 1.
Evidence shows: Claims in README are not backed by traceable evidence. Cross-source corroboration is not performed. Facts and inferences are not clearly separated. Therefore all criteria score 1.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- No implementation details for security features are provided; evaluate with caution.
- Dependency availability is not verified; potential supply chain risks.
- No versioning or changelog provided; maintenance status unclear.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
The last state of the AgentDock framework: agentdock-core/, src/, agents/ and docs/ as they stood before the repository separation. Kept permanently so anything built on the framework stays reachable. Clone or browse at this tag. Context: #354View on GitHub ↗