Docker Agent v1.149.0 – Open‑Source AI Agent Builder and Runtime

TL;DR

Docker Agent v1.149.0 (released Oct 7 2026) provides a Docker‑CLI plug‑in that lets you build, run, and share AI agents entirely from a declarative YAML file, supports dozens of model providers, and enables multi‑agent orchestration without writing code.

---

What Docker Agent v1.149.0 actually does

Docker Agent is a Docker CLI plugin (docker agent) that turns a YAML description into a fully‑featured AI agent runtime. The core ideas are:

• Declarative YAML – Agents, tools, and orchestration are described in a version‑controlled file, making them easy to store in Git and push to OCI registries.
• Multi‑agent architecture – You can define a hierarchy of agents (root, sub‑agents, etc.) that delegate tasks automatically.
• Provider‑agnostic model access – Built‑in support for OpenAI, Anthropic, Gemini, AWS Bedrock, Mistral, xAI, Docker Model Runner, and any MCP (Model Context Protocol) server.
• Rich tool ecosystem – Agents can call built‑in tools (think, todo, memory, RAG) or any custom MCP service, whether local, remote, or container‑based.
• Packaging & sharing – Agents can be packaged as OCI images and pulled/run anywhere, enabling reproducible AI‑driven workflows.

All of this is achieved without writing a single line of Go or Python; the runtime does the heavy lifting.

---

Key Features Highlighted in the Release

Feature	Why it matters
Multi‑agent orchestration	Allows complex problems to be split across specialized agents, reducing prompt‑bloat and improving reliability.
Tool‑first design	Built‑in "think", "todo", and memory tools give agents structured reasoning capabilities beyond raw LLM output.
RAG integration	BM25, embeddings, hybrid search, and reranking let agents retrieve up‑to‑date information from private data sources.
Model‑provider agnosticism	Switch between OpenAI, Anthropic, Gemini, Bedrock, etc., without changing your YAML – just update the model field or provider alias.
OCI‑registry distribution	Agents become portable artifacts, versioned like containers, enabling reproducible AI pipelines across teams.
Telemetry (opt‑out)	Anonymous usage data helps Docker improve the tool while respecting privacy (see the Telemetry page for details).

---

Quick‑Start Guide (Answer‑First)

# 1️⃣ Install – Docker Desktop (4.63+) already includes the plugin
#    or via Homebrew:  brew install docker-agent
#    or download a binary from the GitHub Releases page.

# 2️⃣ Export an API key for your preferred provider
export OPENAI_API_KEY=sk-…   # or ANTHROPIC_API_KEY, GOOGLE_API_KEY, etc.

# 3️⃣ Run the built‑in default agent
docker agent run

# 4️⃣ Run an agent stored in an OCI registry
docker agent run myorg/agent:tag

# 5️⃣ Generate a new agent interactively
docker agent new

# 6️⃣ Run a custom YAML config
cat > agent.yaml <<'EOF'
agents:
  root:
    model: openai/gpt-5-mini
    description: A helpful AI assistant
    instruction: |
      You are a knowledgeable assistant that helps users with various tasks.
    toolsets:
      - type: mcp
        ref: docker:duckduckgo
EOF

docker agent run agent.yaml

The docker agent new wizard helps you scaffold a YAML file, and the docker agent run command launches the runtime inside a lightweight container.

---

Community Reaction on Hacker News

The post attracted 142 comments; several themes emerged:

Orchestration convergence – "The models are converging (directionally), and the rest of us, including all these agent harnesses, are trying to figure out how to stay relevant" (pomsense). This reflects a broader industry trend where multiple projects (Docker Agent, Pullboard, LangChain sandboxes, etc.) are converging on similar orchestration concepts.

"No code required" debate – "How is ‘no code required’ a selling point when we all have agents that can instantly puke out mostly‑correct code?" (woah). The community questions whether declarative YAML truly eliminates code, noting that the underlying toolchain is still code‑heavy.

Coherency over time – "The bigger challenge is handling agent coherency over long periods (drift)… I built a forum‑like system where agents shout to one another and document work" (Olscore). This highlights a limitation of many harnesses: maintaining state and correctness across long‑running workflows.

Security concerns – "If you couldn't find any security‑related info on the linked page" (McScrooge). Users are looking for clearer sandboxing and isolation guarantees.

Comparison with other frameworks – Multiple commenters (blakeashleyjr, maxdo, esafak) noted that Docker Agent feels like another JS‑style framework, and asked why Docker’s name is attached when similar capabilities exist in projects like Modal, e2b, or Cloudflare.

Use‑case clarity – "I'm honestly getting lost between what primary use‑case each of these projects is intended for" (blutoot). The community seeks clearer positioning of Docker Agent relative to Pullboard, kagent.dev, LangChain sandboxes, etc.

Overall, the discussion underscores excitement about a unified, container‑native agent runtime, but also a desire for more mature security, state‑management, and clear differentiation.

---

How Docker Agent Fits Into the AI‑Developer Landscape

1. Container‑native execution – By leveraging Docker’s existing sandboxing model, agents run with the same isolation guarantees as any container, addressing concerns about untrusted LLM code execution.
2. Versioned, reproducible AI pipelines – Storing agents as OCI images aligns AI workflows with existing CI/CD practices, making it easier for enterprises to audit and roll back AI behavior.
3. Extensible toolchain – The MCP protocol lets any service expose tools to agents, turning the runtime into a universal orchestrator for cloud, on‑prem, and edge resources.
4. Open‑source community – With 3.7k stars, 485 forks, and over 100 contributors, Docker Agent benefits from a broad contributor base, which should accelerate feature maturity and security hardening.

---

Getting Involved

• Read the docs – Full documentation lives at https://docker.github.io/docker-agent/.
• Contribute – Follow the contributing guide; the repository uses the same docker agent CLI to build itself (docker agent run ./golang_developer.yaml).
• Join the community – Docker Community Slack (#docker‑agent) is the primary venue for discussion, feature requests, and troubleshooting.

---

Bottom Line

Docker Agent v1.149.0 delivers a production‑ready, container‑native platform for building and sharing AI agents with zero‑code YAML configuration, multi‑agent orchestration, and broad model/provider support. While the community notes that the ecosystem still needs stronger security guarantees and better long‑term state management, the project marks a significant step toward reproducible, portable AI‑driven developer workflows.