LangGraph vs CrewAI: which one?

CrewAI gets you a convincing multi-agent demo with the least code, because it hands you roles and a process model out of the box. LangGraph makes you draw the graph yourself, which is more work on day one and much less work on day ninety, when you need to know exactly why the agent looped four times. Prototype in CrewAI if speed matters; build the thing customers touch in LangGraph if it has real branching.

Is the OpenAI Agents SDK a lock-in risk?

Somewhat, and it's a fair trade for many teams. The SDK is thin and the concepts port, but its best features (hosted file search, web search, computer use) are OpenAI's. If model-switching is a board-level requirement, pick something provider-agnostic. If it isn't, don't pay abstraction costs to avoid a problem you may never have.

What is smolagents good for if it's so small?

Learning, and more production work than its size suggests. It's a barebones library where agents write code as their action space, and you can read the whole thing in an evening. Teams that start there tend to make better framework decisions afterwards, because they know what the abstraction is hiding.

Should I use Google ADK?

If your data and deployment already live in Google Cloud, it's a reasonable default: code-first, Apache-2.0, with the integration story as the selling point. If you're not on GCP, the ecosystem around LangGraph is larger and you'll find answers faster.

Can I use two frameworks together?

You can. You almost always shouldn't. Two orchestration layers means two state models, two failure vocabularies and two sets of upgrade notes, and the boundary between them becomes the place bugs live. The exception is a genuinely separate service with a separate team.

Agent Library · Orchestration

AI agent orchestration frameworks: the six worth your time, and how to choose one

The six open-source agent orchestration frameworks worth evaluating in 2026 are LangGraph, the OpenAI Agents SDK, CrewAI, AutoGen, smolagents and Google ADK. Choose LangGraph for stateful workflows that branch, retry and pause for human review; the OpenAI Agents SDK for the fastest path to a working loop inside one provider; CrewAI for role-based multi-agent prototypes; smolagents to learn the pattern before committing. All six are free and open source.

Free. 6 repositories, no payment. One email address opens the list.

Every framework on this page solves the same small problem. Call a model, read what it wants, run the tool, feed the result back, repeat. That loop is about forty lines. You could write it this afternoon.

What you're actually choosing between is opinions: how state is stored, what happens when step four throws, whether a human can interrupt, and how much of your code you rewrite when you switch models.

The first question I'd ask isn't which framework. It's whether your agent has to survive a restart. If a user can close the app mid-task and expect to come back to it, you need durable state, and that narrows the field to roughly LangGraph plus a fair amount of custom work. If every run is one request and one response, almost anything here is fine, and you should optimise for how fast your team can read the code instead.

The second question is what language your product is already written in. Most of this ecosystem is Python, and a lot of teams end up standing up a second service purely to host the agent: its own container, its own secrets, its own on-call rota. That cost is real and it never appears in the comparison tables. If you're a TypeScript shop, read the TypeScript page before you accept it.

One honest warning on AutoGen. The ideas out of Microsoft Research are strong and the codebase has been substantially rewritten more than once. That's normal for a research-led project and a genuine risk for a production roadmap. Read the migration history before you commit a quarter to it.

This week: write down whether your agent must survive a restart, and what language your product is in. Those two answers pick the framework for you in about five minutes.

Where to send the 6 repositories

Give me an email address and the full list opens on this page straight away, with a copy in your inbox you can search later, plus a note whenever we publish something new on AI and where mobile development is going. No payment, now or ever. That's the only trade.

Orchestration: where the agent actually lives

Pick one. Picking two is how projects die. These are the four we have shipped against, and the honest reason you would choose each.

6 repositories, with the notes on each. Enter your email above to read them.

The master list: all 100 repositories

The free list above is the short version, and every repository on it is in here too. The Gods Repo Library is the whole catalogue: 100 repositories across 12 categories, each with a written note on what it is for and when to reach for it, and it carries the AI optimisation half, inference, serving, quantisation and evaluation, that the free list barely touches. Every figure in it is read from the GitHub API rather than typed by hand, and nothing archived or idle survives the build. Verified 2026-09-23.

Orchestration: where the agent lives19
Coding agents you can point at your own repo6
MCP, tools and the integration layer10
Browser and computer control5
Retrieval: giving the model your data9
Memory and vector stores7
Evaluation, observability and guardrails11
Prompt and output optimisation5
Inference, serving and routing10
Fine-tuning, quantisation and compression7
Getting the data in5
Courses, cookbooks and reference6

$9.99once, for all 100 repositories, sent to your inbox

All 100 repositories with a written note on each, emailed to you
88 of them are beyond the free list above
A permanent link to the live version, refreshed against the GitHub API
No subscription and no account, just the library

Before you buy, the terms of this sale

The full catalogue is delivered to the email address I give, immediately, the moment my payment clears.
Because it is delivered in full straight away, all sales are final: no refunds, no partial refunds, no exceptions, for any reason.
I am asking for that immediate delivery and I give up any cancellation or cooling-off period that would otherwise apply to digital content.
If the email does not reach me, Frenchy Digital will re-send it to my address for as long as I need, and that re-send is the remedy rather than a refund.
The repositories listed are third-party open-source projects. Their availability and their licences are theirs to change, and I will check any licence myself before using it commercially.

I have read and accept the terms above. I understand this purchase is final and no refund is available.

The same terms are in full in our terms and conditions, and you will be asked to accept them once more on the Stripe payment page.

You can change the address at checkout, and that is where it is delivered. Secure checkout by Stripe, and we never see your card.

Also in the library: the Agency CRM Ops Skill

A Claude Code skill that runs an agency’s client operations: a pipeline that only holds real prospects, solicitor detection, a lawful marketing list, meeting recordings filed one folder per client, and inbox cleanup that never touches your favorites. $20 once.

See what the skill does

Tip if it helped. Only if it helped.

The library costs nothing and stays that way. You already have it, whether or not you type anything here. If it saved you an afternoon, name your own amount. It goes toward the time we spend testing these so you don't have to.

Anything from $2. You can still change it on the next screen. Secure checkout by Stripe, and we never see your card.

Our partners

Tools we use or recommend to clients, with the offer each one gives new customers. These are referral links: if you sign up, we may earn a commission at no cost to you. Offers checked September 2026 and set by each company, so confirm on their page.