docs: refresh Deep Agents Code guidance

[deepujain]

Summary

Refreshes the LangChain Deep Agents Code docs so the supported path is clearer for current NemoClaw users.

Changes

• Documents the canonical nemoclaw onboard --agent langchain-deepagents-code path, plus host-side headless nemoclaw <name> agent -n "...".
• Clarifies that dcode uses NemoClaw-managed inference through https://inference.local/v1 and should not receive raw provider keys in .deepagents/.env.
• Adds Tavily to the maintained policy preset docs and explains the Tavily/LangSmith/MCP boundaries for this terminal runtime.
• Calls out that LangChain Deep Agents Code has no dashboard URL, gateway token, forwarded API port, or long-running in-sandbox gateway.

Testing

• npm run docs:check-agent-variants
• npm run docs
• git diff --check
• git diff --cached --check
• Touched-file lint/format check: scoped prek run --files ... and Biome applicability check

Evidence it works

npm run docs completed with Fern reporting 0 errors. The scoped prek run passed the touched-file whitespace, EOF, merge-conflict, large-file, env-var documentation, secret-scan, and related applicable hooks, then exited non-zero because the platform-matrix sync hook rewrote generated tables outside this PR scope. Biome and markdownlint had no MDX files to process. npm run build:cli was not run because this is docs-only and does not change source, manifests, or runtime behavior.

Fixes #5858

Signed-off-by: Deepak Jain deepujain@gmail.com

Summary by CodeRabbit

• Documentation
  • Clarified onboarding and usage for the Deep Agents Code runtime, including scripted setup, headless task execution, and how to connect after onboarding.
  • Added notes explaining that dashboard, gateway, and in-sandbox API guidance does not apply to this runtime.
  • Expanded network-policy docs to include Tavily web search access and when to opt in.
  • Strengthened security guidance around keeping provider and search credentials out of project files and using the managed runtime flow.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

This repository limits contributors to 10 open pull requests. Please close or merge existing PRs before opening new ones.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 7abc74cf-09dd-4b75-a64b-e1f809d104d7

📥 Commits

Reviewing files that changed from the base of the PR and between c1c7c1d and 28802fa.

📒 Files selected for processing (7)

• docs/get-started/quickstart-langchain-deepagents-code.mdx
• docs/inference/inference-options.mdx
• docs/network-policy/customize-network-policy.mdx
• docs/network-policy/integration-policy-examples.mdx
• docs/reference/commands.mdx
• docs/reference/troubleshooting.mdx
• docs/security/credential-storage.mdx

---

📝 Walkthrough

Walkthrough

Updated LangChain Deep Agents Code docs to cover onboarding, terminal-runtime commands, optional Tavily policy setup, and managed inference and credential storage behavior.

Changes

LangChain Deep Agents Code docs refresh

Layer / File(s)	Summary
Onboarding and runtime usage docs/get-started/quickstart-langchain-deepagents-code.mdx, docs/reference/*.mdx	Adds non-interactive onboarding, host-side agent -n usage, terminal-runtime connect and status notes, and troubleshooting text stating that dashboard-oriented guidance does not apply.
Tavily policy guidance docs/get-started/quickstart-langchain-deepagents-code.mdx, docs/network-policy/*.mdx	Adds quickstart notes for optional web search, includes the tavily preset in policy tables, and adds a Deep Agents Code Tavily section that keeps API keys at runtime and separates LangSmith/MCP guidance.
Inference and credential handling docs/inference/inference-options.mdx, docs/security/credential-storage.mdx	Describes /sandbox/.deepagents/config.toml targeting https://inference.local/v1 with a placeholder key and expands credential-storage guidance for sandbox .env and committed files.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• NVIDIA/NemoClaw#5651: Updates the same LangChain Deep Agents Code and Tavily documentation path around opt-in web search and runtime credential handling.
• NVIDIA/NemoClaw#5815: Adds terminal-runtime dispatch behavior for langchain-deepagents-code that these docs describe for agent -n, connect, and dcode.

Suggested labels

area: docs, area: policy, integration: dcode

Suggested reviewers

• cv
• jyaunches
• prekshivyas

Poem

A bunny hopped through docs so bright,
With dcode steps and policies in sight.
Tavily waits at the gate, not by default,
Secrets stay tucked in the burrow vault.
Thump-thump — the quickstart feels just right.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}