Index of every page on this site: llms.txt
All skills
google avatar

/agents-cli-langchain

@2fa159a
by googlegoogle/agents-cli6.4k stars
738
Run it yourself

Use when working in this project — adding tools, editing the agent, running, evaluating, serving or deploying it — or when guidance mentions ADK, LlmAgent, google.adk, adk web, or an ADK runner. This project is LangChain/LangGraph scaffolded by agents-cli, so ADK-specific instructions do not apply.

  • 3 files
  • 15.5 KB
  • Updated 4 days ago
  • GitHub
Use this Skill: https://skilld.dev/gh/google/agents-cli/agents-cli-langchain

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈81 tokens for metadata: the name and description. ≈1.1k when used: this file. ≈2.8k more on demand in 2 files.

Description uses 4.0% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈322 characters, or 4.0%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

LangChain project (agents-cli)

The agent is a compiled LangGraph graph exported as root_agent from app/agent.py. There is no google.adk dependency and no ADK runner. Other google-agents-cli-* skills assume ADK; where they describe the agent itself, this skill wins.

Experimental, and Agent Runtime is degraded. Deploy to cloud_run or gke. On agent_runtime the app serves, but publish gemini is refused, the Console playground cannot invoke it, and Console sessions/traces stay empty: all three want the ADK reasoning_engine routes this project does not serve. Say so before recommending it.

What ADK guidance maps to here

ADK guidance This project
LlmAgent, Agent, google.adk.tools langchain.agents.create_agent, plain Python functions as tools, or any compiled StateGraph
adk web, adk run agents-cli playground (runs langgraph dev)
ADK runner behind agents-cli run agents-cli run invokes the graph in-process
agents-cli eval dataset synthesize, eval optimize Unavailable: both drive the agent through ADK. The command says so and exits
Add an LlmAgent in app/agent.py Change the graph in app/agent.py; keep the name root_agent

The contract

Keep these two, whatever you build inside them:

  • app/agent.py exports root_agent, a compiled graph with messages state. Callers only use root_agent.invoke({"messages": [...]}) and root_agent.astream(stream_mode="messages").
  • app/fast_api_app.py exposes app. Every deployment target runs uvicorn app.fast_api_app:app.

Adding a tool means writing a typed function with a docstring and passing it in tools=[...]. Switching frameworks (LangGraph StateGraph, deepagents.create_deep_agent) means rewriting app/agent.py only. Pre-1.0 LangChain (LCEL chains, AgentExecutor) is not supported: not compiled graphs.

Commands

agents-cli install                  # uv sync
agents-cli playground               # langgraph dev, port 8080
agents-cli run "hello"              # invoke the graph in-process
agents-cli eval generate --dataset tests/eval/datasets/basic-dataset.json -o tests/eval/output/
agents-cli eval grade --traces tests/eval/output/<dataset>.json --config tests/eval/eval_config.yaml
agents-cli deploy                   # unchanged
agents-cli scaffold enhance -d cloud_run --cicd-runner github_actions   # add infra later

playground, run and eval generate are overridden by agents-cli-extension.yaml at the project root. Prefix any command with AGENTS_CLI_DISABLE_OVERRIDES=1 to reach the built-in instead.

Serving & Persistence

A2A only: JSON-RPC at POST /a2a/app, card at /a2a/app/.well-known/agent-card.json, health at /health. Token streaming comes from astream(stream_mode="messages").

app/app_utils/services.py wires both a short-term thread checkpointer (keyed by A2A context_id as thread_id) and a long-term BaseStore (reachable in tools via from langgraph.prebuilt import InjectedStore) onto root_agent at server startup:

  • With --session-type cloud_sql on Cloud Run or GKE (INSTANCE_CONNECTION_NAME + DB_PASS, or SESSION_DB_URI), it uses AsyncPostgresSaver and AsyncPostgresStore sharing one connection pool.
  • Without DB credentials, it falls back to InMemorySaver and InMemoryStore.
  • Leave root_agent in app/agent.py without a custom checkpointer or store at module import time so langgraph dev (agents-cli playground) can attach its own runtime persistence.

Common mistakes

  • Renaming root_agent or app, which breaks run, eval and deploy.
  • Reaching for eval dataset synthesize or eval optimize: they need ADK. Write cases into tests/eval/datasets/ and use eval generate + eval grade.
  • Expecting /run_sse or ADK session routes; this server serves A2A.
  • Running agents-cli run --url ... against a deployed agent without AGENTS_CLI_DISABLE_OVERRIDES=1, which invokes the local graph instead.
  • run and eval generate call Gemini on Gemini platform with ADC, so they need GOOGLE_CLOUD_PROJECT and credentials, or GOOGLE_API_KEY / GEMINI_API_KEY in .env.

References

  • references/langchain.md — framework contract and per-command detail.
  • references/samples.md — agents worth copying from, by shape.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at 2fa159a. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 days ago.

Activeupdated 4 days ago

README badge

README badge for google/agents-cli/agents-cli-langchain