CodeWords vs Gumloop: which one should you use?

Short answer: choose Gumloop if you want AI agents your team can chat with and hand tasks to in Slack, Teams or email, rolled out across the company with IT controls. Choose CodeWords if you have a recurring job you can describe: Cody builds it into a workflow that runs on its own in your tools, and you describe each change.

See it with a real job: the Reddit Lead Miner template. It finds high-intent B2B leads on Reddit, qualifies them, and adds them to your Notion CRM.

Gumloop and CodeWords both start from plain language now. Describe what you need, and either one will connect your apps and set it up. The difference is what you are left with. Gumloop gives you an agent: each time it runs, it reads the prompt you configured and works out what to do then.¹ CodeWords gives you a workflow: Cody works out the steps when you describe the job, builds and deploys them, and that workflow is what runs until you describe a change.

• CodeWords: describe the job once in plain language. Cody builds it, connects your tools, deploys it, and keeps it running. When it needs to change, you describe the change.
• Gumloop: in its own words, "the multiplayer AI agent builder".² Anyone at a company can build an agent that uses their tools, while IT controls access. Its older visual flowchart builder is still available as Workflows (Legacy).³

TL;DR

• A job versus a teammate. Gumloop is built around agents that people chat with and hand tasks to. CodeWords is built around one recurring job that runs on its own.
• An agent decides on each run; a workflow runs what you reviewed. A triggered Gumloop agent handles its prompt "exactly as if you had typed it in the chat". A CodeWords workflow runs the steps Cody built and you checked.
• Both change in chat. Tell a Gumloop agent to update its instructions, or tell Cody what is different and it updates and redeploys the workflow.
• Gumloop is the better choice if your team wants an assistant in Slack, Teams or email, or you are rolling AI agents out across a company and need IT controls over who uses what.

How CodeWords is different

1. The steps are worked out once, not on every run

Gumloop's agents can run on their own, on a schedule, a webhook or an app event.¹ Each run starts the agent again: it gets the prompt you configured and decides which tools to call from there. That is the strength of an agent, because it can take on a task nobody planned for.

On CodeWords, Cody plans the job when you describe it, writes and tests the code behind it, connects your tools and deploys it. What runs on Tuesday is the workflow you described and reviewed. When the job changes, you describe the change and Cody updates the workflow. See how that works from a single sentence.

2. Built around the job, not the people using it

Gumloop is designed for one person to build an agent and the rest of the company to use it, in Slack, Microsoft Teams, email or a hosted page, with access controls for IT.⁴

CodeWords is built around the job itself. You describe a recurring process with a named input, a destination and an output someone depends on. What you get back is a workflow that runs on its own, acts in your tools, and keeps a record of what it did.

3. You own the code, on every plan

Everything Cody builds is real code, and it is yours on every plan including the free one. You never have to open it, but it is not a black box you are locked out of. A Gumloop agent is a configuration of instructions, skills and connectors that runs on Gumloop.

If you have a recurring job you would rather describe once than hand to an agent each time, start building with CodeWords for free.

Comparing CodeWords and Gumloop

Both platforms start from plain language, both connect your tools, and both can run on a schedule or a trigger. Here is where they actually differ.

Core features

	CodeWords	Gumloop
Who it's for	People who can describe a job and judge its output	Anyone at a company, with IT controlling access2
How you build	Describe the job to Cody	Tell a new agent what it should take care of; it connects apps and writes its own instructions5
What you hold afterwards	A deployed workflow	An agent: instructions, skills, connectors and triggers4
What runs each time	The steps Cody built and you reviewed	The agent, which decides what to do on each run1
Making a change	Describe the change in the same chat	Tell the agent; it can edit its own instructions, skills and triggers5
Fixing something that breaks	Debug mode explains what went wrong in plain language — ask Cody to fix it, and it does	Correct the agent in chat; it also reviews its own runs and suggests fixes2
Where people use it	It runs on its own in your tools	Chat, Slack, Microsoft Teams, email, a hosted page, the API or MCP4
Running unattended	The default, from the first build	App events, webhooks, schedules, or one-time runs1
Integrations	3,000+ integrations, plus bring-your-own API keys	250+ MCP servers6
Visual flowchart builder	No, by design	Yes, as Workflows (Legacy)3

Pricing comparison

Both platforms meter usage rather than charging a flat fee, so compare them against what you actually plan to run.

CodeWords

• Free — $0/month, 100 workflow runs, unlimited projects
• Pro — $39/month, on-demand credit top-ups
• Business — $100/month, priority support
• Custom — contact sales
• Runs on one credit pool that covers building, running, and connecting your tools

See the full CodeWords pricing for current plans.

Gumloop³

• Pro — from $37/month, 20,000 credits a month, unlimited seats and agents, 14-day free trial
• Enterprise — custom pricing, contact sales
• No free plan: the trial needs a card and rolls into Pro⁷
• Each run is billed for the model tokens it uses at the provider's price, plus credits for each tool call and each active minute, plus an 8% orchestration fee (16% with your own model key)⁷

When to choose CodeWords vs Gumloop

Choose CodeWords if you have a recurring job you can describe precisely and you want it to run on its own, in your tools, with a record of what it did. Choose Gumloop if what your team wants is an AI agent to chat with and hand tasks to, rolled out across the company with IT controls.

Recommendations by team type

• Someone with a recurring job and no developer to hand it to who wants it live this week — CodeWords.
• A marketing or ops team with a process that should run on its own and that they would rather describe once than delegate each time — CodeWords.
• An agency running similar work for several clients who would rather adjust a description than reconfigure an agent for each one — CodeWords.
• A team that wants an AI assistant in Slack, Teams or email that people hand tasks to — Gumloop.
• A company rolling AI agents out to every team, with IT controls over access — Gumloop.
• Ad hoc tasks people ask for in chat, different each time — Gumloop.

Frequently asked questions

Can I move my existing Gumloop agents or flows into CodeWords? Yes, for recurring jobs. Describe what the agent or flow does, or paste a screenshot of a legacy flow, into Cody. It rebuilds the same job as a working workflow.

Gumloop builds agents from a description too. Is that the same thing? The first step looks similar: both start from plain language. The difference is what is left behind. Gumloop gives you an agent that decides what to do each time it runs. CodeWords gives you a deployed workflow that runs the steps you reviewed, and changes when you describe the change. See what an AI agent actually is for the underlying concept.

Which is cheaper, CodeWords or Gumloop? It depends on what you build and how often it runs. Both meter usage, so compare each against your real usage rather than the headline plan price.

Describe the work you keep repeating and see what Cody builds. Try CodeWords free. Or start from the Reddit Lead Miner template.

Footnotes

1. Gumloop docs, Agent triggers, checked 7 October 2026. ↩ ↩² ↩³ ↩⁴

2. Gumloop, homepage, checked 7 October 2026. ↩ ↩² ↩³

3. Gumloop, Pricing, checked 7 October 2026. ↩ ↩² ↩³

4. Gumloop docs, Agents, checked 7 October 2026. ↩ ↩² ↩³

5. Gumloop, Changelog, checked 7 October 2026. ↩ ↩²

6. Gumloop, MCP servers, checked 7 October 2026. ↩

7. Gumloop docs, Credits, checked 7 October 2026. ↩ ↩²