experimental(agent): Add coding agent workflow adapters

.gitignore

@@ -1,4 +1,5 @@
 ###### Place new entries directly below this line! ######
+node_modules/
 /.codex
 
 # auto-generated chainlit stuff

examples/README.md

@@ -35,6 +35,7 @@ The examples in the NeMo-Agent-Toolkit-Examples repository are community contrib
 - **[NeMo Agent Toolkit Components](#nemo-agent-toolkit-components)**
   - [Agents](#agents)
   - [Advanced Agents](#advanced-agents)
+  - [Experimental](#experimental)
   - [Configuration](#configuration)
   - [Control Flow](#control-flow)
   - [Custom Functions](#custom-functions)
@@ -104,6 +105,13 @@ _Additional information can be found in the [Agents README](./agents/README.md).
 - **[`alert_triage_agent`](advanced_agents/alert_triage_agent/README.md)**: Production-ready intelligent alert triage system using LangGraph that automates system monitoring diagnostics with tools for hardware checks, network connectivity, performance analysis, and generates structured triage reports with root cause categorization **[🟨 Intermediate]**
 - **[`vulnerability_analysis_blueprint`](advanced_agents/vulnerability_analysis_blueprint/README.md)**: Blueprint documentation for vulnerability analysis agents **[🛑 Advanced]**
 
+### Experimental
+- **[`claude_code_agent_adapter`](experimental/claude_code_agent_adapter/README.md)**: Experimental primitive agent workflow backed by the Claude Code Agent SDK with a conservative local smoke-test config. **[🛑 Advanced]**
+- **[`codex_agent_adapter`](experimental/codex_agent_adapter/README.md)**: Experimental primitive agent workflow backed by Codex SDK in read-only mode. **[🛑 Advanced]**
+- **[`cursor_agent_adapter`](experimental/cursor_agent_adapter/README.md)**: Experimental primitive agent workflow backed by Cursor Agent CLI in plan mode. **[🛑 Advanced]**
+- **[`openclaw_agent_adapter`](experimental/openclaw_agent_adapter/README.md)**: Experimental primitive agent workflow backed by OpenClaw CLI one-shot runs. **[🛑 Advanced]**
+- **[`hermes_agent_adapter`](experimental/hermes_agent_adapter/README.md)**: Experimental primitive agent workflow backed by Hermes Agent CLI one-shot mode. **[🛑 Advanced]**
+
 ### Configuration
 - **[`config_inheritance`](config_inheritance/README.md)**: Use YAML configuration inheritance in the NeMo Agent Toolkit to reduce duplication across similar configuration files **[🟢 Beginner]**
 

examples/experimental/claude_code_agent_adapter/README.md

@@ -0,0 +1,104 @@
+<!--
+SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+SPDX-License-Identifier: Apache-2.0
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+-->
+
+# Claude Code Adapter
+
+This experimental NVIDIA NeMo Agent Toolkit example prototypes a primitive agent workflow type backed by the Claude Code Agent SDK.
+
+The adapter is intentionally conservative:
+
+- It accepts the same `ChatRequestOrMessage` input shape used by built-in workflow agents.
+- Its config subclasses NAT's `AgentBaseConfig`; `llm_name` is optional because Claude Code Agent SDK manages model selection through `model`.
+- Operational controls live in YAML config rather than an NAT `llms:` block.
+- The default SDK permission mode is `plan`.
+
+## Installation And Setup
+
+If you have not already done so, follow the instructions in the [Install Guide](../../../docs/source/get-started/installation.md#install-from-source) to create the development environment and install NeMo Agent Toolkit.
+
+### Install This Workflow
+
+From the root directory of the NeMo Agent Toolkit library, run:
+
+```bash
+uv pip install -e examples/experimental/claude_code_agent_adapter
+```
+
+## Run The Workflow
+
+For local end-to-end testing, authenticate Claude Code in the same shell or user profile that launches `nat`:
+
+```bash
+claude auth status
+claude auth login
+```
+
+If `claude auth status` shows an active NVIDIA SSO-backed Claude login, run the live config:
+
+```bash
+nat run \
+  --config_file examples/experimental/claude_code_agent_adapter/configs/config.yml \
+  --input "Inspect the experimental Claude Code agent adapter and summarize how it registers with NAT. Do not edit files."
+```
+
+`configs/config.yml` keeps `permission_mode: plan`, uses the `sonnet` model alias, caps `max_budget_usd`, and denies write-oriented tools for the first live smoke test.
+
+## Direct Workflow
+
+`configs/config.yml` wires the adapter directly as the workflow:
+
+```yaml
+workflow:
+  _type: claude_code_agent
+  permission_mode: plan
+  model: sonnet
+  working_directory: .
+  setting_sources: [project]
+  max_turns: 10
+  max_budget_usd: 1.00
+  disallowed_tools:
+    - Bash
+    - Edit
+    - MultiEdit
+    - NotebookEdit
+    - Write
+```
+
+## Configurable Options
+
+- `working_directory`
+- `llm_name` (accepted for NAT agent config consistency, unused by the SDK agent)
+- `permission_mode`
+- `model`
+- `append_system_prompt`
+- `allowed_tools`
+- `disallowed_tools`
+- `setting_sources`
+- `additional_directories`
+- `max_turns`
+- `max_budget_usd`
+- `timeout_seconds`
+- `max_output_chars`
+- `max_history`
+
+## Authentication
+
+The adapter does not require credentials in YAML. In live mode, the Claude Agent SDK follows Claude Code authentication in the launching environment. For local testing, an NVIDIA SSO-backed Claude Code login can be used if `claude auth status` reports an active login in the same shell/user profile. Stale `ANTHROPIC_API_KEY` or `ANTHROPIC_AUTH_TOKEN` environment variables can take precedence over that login.
+
+## Notes
+
+Live execution should stay opt-in until permission, sandboxing, credentials, and cost controls are reviewed. The current agent implementation uses `claude_agent_sdk.query()` and `ClaudeAgentOptions`.

examples/experimental/claude_code_agent_adapter/configs/config.yml

@@ -0,0 +1,19 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+workflow:
+  _type: claude_code_agent
+  permission_mode: plan
+  model: sonnet
+  working_directory: .
+  setting_sources: [project]
+  max_turns: 10
+  max_budget_usd: 1.00
+  timeout_seconds: 120
+  max_output_chars: 12000
+  disallowed_tools:
+    - Bash
+    - Edit
+    - MultiEdit
+    - NotebookEdit
+    - Write

examples/experimental/claude_code_agent_adapter/pyproject.toml

@@ -0,0 +1,45 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+[build-system]
+build-backend = "setuptools.build_meta"
+requires = ["setuptools>=64", "setuptools-scm>=8", "setuptools_dynamic_dependencies>=1.0.0"]
+
+[tool.setuptools_scm]
+git_describe_command = "git describe --long --first-parent"
+root = "../../.."
+
+[tool.setuptools.packages.find]
+where = ["src"]
+
+[project]
+name = "nat_claude_code_agent_adapter"
+dynamic = ["version", "dependencies"]
+requires-python = ">=3.11,<3.14"
+description = "Experimental Claude Code agent for the NeMo Agent Toolkit"
+keywords = ["ai", "agents", "coding-agents"]
+classifiers = ["Programming Language :: Python"]
+
+[tool.setuptools_dynamic_dependencies]
+dependencies = [
+  "claude-agent-sdk>=0.1.60",
+  "nvidia-nat[langchain] == {version}",
+]
+
+[tool.uv.sources]
+nvidia-nat = { path = "../../..", editable = true }
+
+[project.entry-points.'nat.components']
+nat_claude_code_agent_adapter = "nat_claude_code_agent_adapter.register"

examples/experimental/claude_code_agent_adapter/src/nat_claude_code_agent_adapter/__init__.py

@@ -0,0 +1,15 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Experimental Claude Code adapter example."""