Setup Guide

Run omp with a free coding model

Connect omp to OpenRouter and start with Poolside: Laguna S 2.1 (free). Your first agentic coding run costs $0 in model tokens. The free-model catalog refreshes every 24 hours, and every setup on this page keeps the explicit :free suffix.

terminal

$ omp

model: openrouter/poolside/laguna-s-2.1:free · $0 model tokens

Free catalog · 24-hour refresh

Pick the model omp will use

Start with Laguna S 2.1 for agentic coding. Choose another model and the terminal example, YAML, manual commands, and Claude Code prompt update together.

Every option below currently has explicit :free pricing, text output, tools, and tool choice. OpenRouter can change availability and rate limits. Review the provider’s data policy before sending private code.

Model

Context

Use it for

Published evidence

Setup

NVIDIA: Nemotron 3 Ultra (free)
1M tokens
Long-context reasoning, orchestration, and coding-agent experiments.
OpenRouter lists a 1,000,000-token context window, reasoning controls, and tool calling. Source
NVIDIA: Nemotron 3.5 Lightning (free)
1M tokens
High-throughput agent experiments and smaller coding tasks.
OpenRouter lists a 1,000,000-token context window, reasoning controls, and tool calling. Source
Dots Studio: Dots3-Note Preview (free)
512k tokens
Open the model card before choosing it for a repository.
No GraniteAI coding recommendation has been recorded for this model yet. Source
Apodex: Apodex 1.1 Mini (free)
262k tokens
Open the model card before choosing it for a repository.
No GraniteAI coding recommendation has been recorded for this model yet. Source
Google: Gemma 4 26B A4B (free)
262k tokens
Lighter multimodal instruction following with tool use.
OpenRouter lists tools, reasoning controls, and a 262,144-token context window. Source
Google: Gemma 4 31B (free)
262k tokens
General multimodal instruction following with tool use.
OpenRouter lists tools, reasoning controls, and a 262,144-token context window. Source
NVIDIA: Nemotron 3 Super (free)
262k tokens
Multi-agent experiments and structured repository work.
OpenRouter lists tool calling, structured output, and a 262,144-token context window. Source
Poolside: Laguna S 2.1 (free)
Recommended
Poolside says free inputs and outputs may be used to improve its models.
262k tokens
Agentic coding, terminal work, and multi-step repository changes.
70.2% Terminal-Bench 2.1 and 40.4% DeepSWE, published on the OpenRouter model card. Source
Poolside: Laguna XS 2.1 (free)
262k tokens
Smaller agentic coding tasks where latency matters more than maximum capability.
Coding-focused Poolside model with tool calling and a 262,144-token context window. Source
Cohere: North Mini Code (free)
256k tokens
Fast routine coding tasks and small tool-driven changes.
Coding-focused model with tool calling and a 256,000-token context window. Source
NVIDIA: Nemotron 3 Nano Omni (free)
256k tokens
Open the model card before choosing it for a repository.
No GraniteAI coding recommendation has been recorded for this model yet. Source
LiquidAI: LFM2.5-2.6B (free)
66k tokens
Open the model card before choosing it for a repository.
No GraniteAI coding recommendation has been recorded for this model yet. Source

Catalog read Oct 11, 2026, 07:57 PM UTC. Availability and technical fields come from the public OpenRouter catalog. Fit notes and evidence links were reviewed by GraniteAI on Aug 27, 2026. They are not equivalence rankings.

Steps

~10

Minutes

Model tokens

24h

Catalog refresh

For Claude Code users

Give Claude Code the setup job

Paste the prompt on the right into Claude Code. It opens OpenRouter for sign-up, installs omp, registers all 12 free models above, and runs one :free test call. A new account calls them with no credit card and no credits.

The guardrails are part of the prompt. No paid fallback. No API key in chat. No credentials typed for you. Run omp, press /model, and the whole free list is yours to switch between. The manual path below exposes every command.

paste this into Claude Code

Set up the omp coding agent on this machine and register every free OpenRouter coding model.

OpenRouter is one account and one API key that reaches models from many providers. A model id ending in :free costs $0 per token.

Safety rules:
- Never print my full API key after I confirm it.
- Do not ask me to paste the key into chat or save it in a project file.
- Ask me before any action that could create a charge. A paid model id needs my approval.
- Keep the :free suffix on every model id and command.

1. Read the free catalog.
   Fetch https://openrouter.ai/api/v1/models?max_price=0&output_modalities=text without authentication.
   Keep every model whose id ends in :free, whose prompt and completion prices are both numeric zero, and whose supported_parameters include tools and tool_choice.
   Tell me how many you found. Drop any model in the roster below that is gone from that list. Add any new one that qualifies.

2. Get me an OpenRouter account.
   Open https://openrouter.ai in a browser. If I am already signed in, go to step 3.
   Otherwise open https://openrouter.ai/sign-up and stop there. Tell me to create the account myself in that window.
   Never type my email, my password, or a verification code. Wait for me to say I am signed in.
   A new account needs no credit card and no credits to call a :free model.

3. Create the API key.
   Open https://openrouter.ai/settings/keys. Tell me to select Create Key and name it omp-free.
   Do not read the key, copy it into chat, or write it to a project file.
   Give me the command for my platform with <key> left as a placeholder:
   macOS/Linux: export OPENROUTER_API_KEY=<key>
   Windows PowerShell: [System.Environment]::SetEnvironmentVariable('OPENROUTER_API_KEY','<key>','User')
   On macOS/Linux, tell me to add that export line to ~/.zshrc or ~/.bashrc so it survives a new terminal.
   On Windows, tell me to open a new PowerShell so the user variable loads.
   Wait for me to confirm the key is set. Then continue.

4. Install omp.
   macOS/Linux: curl -fsSL https://omp.sh/install | sh
   Windows PowerShell: irm https://omp.sh/install.ps1 | iex
   Verify with "omp --version". If the shell cannot find omp, tell me to open a new terminal and run it again. The installer edits PATH and an open shell keeps the old one.

5. Register every free model.
   Write ~/.omp/agent/models.yml (Windows: %USERPROFILE%\.omp\agent\models.yml).
   If the file exists, merge into it. Keep every provider, model, and setting already there. Add these under providers.openrouter.models and replace only an entry with the same id.
   Use this shape for each entry:

   providers:
     openrouter:
       models:
         - id: poolside/laguna-s-2.1:free
           name: "Poolside: Laguna S 2.1 (free)"
           reasoning: true
           input: [text]
           contextWindow: 262144
           maxTokens: 32768
           cost:
             input: 0
             output: 0
             cacheRead: 0
             cacheWrite: 0
           compat:
             thinkingFormat: openrouter

   Write one entry per row below. Every field value comes from this roster:

   poolside/laguna-s-2.1:free | Poolside: Laguna S 2.1 (free) | contextWindow 262144 | maxTokens 32768 | reasoning true
   nvidia/nemotron-3-ultra-550b-a55b:free | NVIDIA: Nemotron 3 Ultra (free) | contextWindow 1000000 | maxTokens 65536 | reasoning true
   nvidia/nemotron-3.5-lightning:free | NVIDIA: Nemotron 3.5 Lightning (free) | contextWindow 1000000 | maxTokens 65536 | reasoning true
   dots-studio/dots-3-note-preview:free | Dots Studio: Dots3-Note Preview (free) | contextWindow 512000 | maxTokens 460800 | reasoning true
   apodex/apodex-1.1-mini:free | Apodex: Apodex 1.1 Mini (free) | contextWindow 262144 | maxTokens 235929 | reasoning true
   google/gemma-4-26b-a4b-it:free | Google: Gemma 4 26B A4B (free) | contextWindow 262144 | maxTokens 32768 | reasoning true
   google/gemma-4-31b-it:free | Google: Gemma 4 31B (free) | contextWindow 262144 | maxTokens 32768 | reasoning true
   nvidia/nemotron-3-super-120b-a12b:free | NVIDIA: Nemotron 3 Super (free) | contextWindow 262144 | maxTokens 235929 | reasoning true
   poolside/laguna-xs-2.1:free | Poolside: Laguna XS 2.1 (free) | contextWindow 262144 | maxTokens 32768 | reasoning true
   cohere/north-mini-code:free | Cohere: North Mini Code (free) | contextWindow 256000 | maxTokens 64000 | reasoning true
   nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free | NVIDIA: Nemotron 3 Nano Omni (free) | contextWindow 256000 | maxTokens 65536 | reasoning true
   liquid/lfm-2.5-2.6b:free | LiquidAI: LFM2.5-2.6B (free) | contextWindow 65536 | maxTokens 8192 | reasoning true

   Two spaces per indent level. Never a tab.

6. Verify. No credits needed.
   Run "omp models openrouter". Every id from step 5 appears in the list.
   Then run: omp -p --model openrouter/poolside/laguna-s-2.1:free "Reply with only OK"
   Expect exactly OK. That call costs $0 and draws on no balance.
   OpenRouter allows 20 requests per minute on free models, and 50 requests per day for an account that has never bought credits. The daily cap rises to 1000 after a 10 dollar credit purchase. If a call returns 429, tell me which cap it hit.
   If a provider reports a model as unavailable, remove that entry and tell me which one. Never drop the :free suffix and never retry against a paid id.

7. Hand off.
   Tell me to open a project, run "omp", and type /model.
   Tell me how many free models are now in that list, and set Poolside: Laguna S 2.1 (free) as the Default role.
   Tell me I can switch to any other model in that list with /model at no cost.
   Tell me to check a provider's data policy before sending private code.

The working setup

Three pieces. One free coding run.

omp gives the model a repository, tools, and a terminal workflow. OpenRouter gives omp one API key across many providers. The selected :free model does the reasoning and coding without a per-token charge.

This page starts with Poolside: Laguna S 2.1 (free). Agentic coding, terminal work, and multi-step repository changes. 70.2% Terminal-Bench 2.1 and 40.4% DeepSWE, published on the OpenRouter model card.

Free endpoints are made for experiments and personal learning. OpenRouter allows 20 requests per minute, and 50 requests per day on an account that has never bought credits. That daily cap rises to 1000 after a 10 dollar purchase. Endpoints can also disappear and may use different data policies. Read the policy before a private repository. Then switch models from the live list without rewriting the setup by hand.

omp

The coding harness. It reads the repository, edits files, runs commands, and keeps the work inside a terminal session you control.

OpenRouter

The provider layer. One account and one key expose the current free catalog. Explicit :free ids prevent a quiet move onto paid inference.

Poolside: Laguna S 2.1 (free)

Agentic coding, terminal work, and multi-step repository changes. 262k-token context. $0 prompt and completion pricing when this catalog was read.

Step by step

The manual guide

Pick your system. Each step is short on purpose. Commands in the dark boxes have a copy button.

01
Install omp
Open Terminal.
curl -fsSL https://omp.sh/install | sh
Run omp --version. A version number confirms the install.
If Terminal cannot find omp, open a new terminal window. Run omp --version again.
Get an OpenRouter API key
Open openrouter.ai in your browser. Make an account if you do not have one.
Go to Settings, then Keys. Select Create Key. Name it omp-free.
Copy the key into the local command in the next step. Do not paste it into chat or a project file.
03
Give omp the key
In Terminal, run this command. Use your key from step 2:
export OPENROUTER_API_KEY=sk-or-v1-...
This export lasts for this terminal session only.
To keep the key, add the same export line to your ~/.zshrc file. Then run source ~/.zshrc.

Register the model

Create the file ~/.omp/agent/models.yml. Paste this text. Save the file:

providers:
  openrouter:
    models:
      - id: poolside/laguna-s-2.1:free
        name: "Poolside: Laguna S 2.1 (free)"
        reasoning: true
        input: [text]
        contextWindow: 262144
        maxTokens: 32768
        cost:
          input: 0
          output: 0
          cacheRead: 0
          cacheWrite: 0
        compat:
          thinkingFormat: openrouter

05
Check that it works
Run omp models openrouter. Poolside: Laguna S 2.1 (free) appears in the list.
Then run this test. The reply must be exactly OK:
omp -p --model openrouter/poolside/laguna-s-2.1:free "Reply with only OK"
Make it your default
Run omp inside one of your projects.
Type /model and press Enter.
Find Poolside: Laguna S 2.1 (free) under openrouter. Assign it to Default.
New sessions now start with Poolside: Laguna S 2.1 (free).

When something fights back

Four problems, four fixes

The shell cannot find omp

The installer adds omp to your PATH, but an open terminal keeps the old PATH. Close the terminal, open a new one, and try again.

models.yml validation failed

Almost always indentation. YAML needs two spaces per level, never tabs. Keep the id line exactly poolside/laguna-s-2.1:free.

Poolside: Laguna S 2.1 (free) does not appear in /model

Check two things. The file must sit at ~/.omp/agent/models.yml, and OPENROUTER_API_KEY must be set in the terminal you launched omp from. Run omp models openrouter to see what loaded.

The free model is rate-limited or unavailable

A 429 means a cap, not a bill. OpenRouter allows 20 requests per minute, and 50 per day on an account that has never bought credits. Wait for the window to reset, or pick another explicit :free model from the list above. Never remove the suffix to force a paid retry.

The paid part

Here is what we sell.

Everything above this line is free and stays free. Below it is what we sell. The prices are the prices, and the two subscriptions carry 30 days money back on the first month.

The readiness audit, done for you

Send us access to your repo. We run the nine-lane review, change nothing, and hand back the report with a 30 minute call on the three gaps to close first.

Who it is for: You have a product you charge for, or want to, and you would rather read the verdict than run the agents.
Price: $349One week
What you get: The report at docs/commercial-readiness.md: a verdict, a scorecard, every gap with the file to open and how to verify it
Legal, security, billing, support, marketing, operations and the rest, checked at once
A 30 minute call on the three gaps to close first
Nothing in your repo changed. The workflow prompt is yours to run afterward
The facts: The review is the one in the video. On our own product it found 87 gaps. The overnight run that followed was 148 commits and one pull request.
Our guarantee: If the review finds no gap worth your time, you pay nothing.

The first run, with Jason

We run the review on your repo together, on a call, and you leave with the report in your repo and the three gaps to close first.

Who it is for: You have the kit and a repo, and you would rather do the first run with someone who has done it.
Price: $9960 minutes
What you get: The review running on your machine, in your Claude Code
The verdict read together, and the three gaps picked
The workflow prompt set up so you can run it that night
The call recording
The facts: New. You would be one of the first. The run itself is the one in the video: 190 agents, 148 commits, one pull request.
Our guarantee: If the review will not run on your setup during the call, there is no charge.

Mica Social Media Analyzer

Your Instagram, answered by the ChatGPT or Claude you already use. Connect the account once, then ask in plain words what worked and what to post next.

Who it is for: You post on Instagram and you are done guessing from the Insights tab.
Price: $29/ month
What you get: Which Reels held attention, ranked, without opening Insights
Readings at 1, 24 and 72 hours after every post, so you see what happened while it was moving
Next captions written in the style of your most saved posts
Read only. It never posts or changes anything
The facts: 7 days free with a card on file. About a dollar a day after that.
Our guarantee: 30 days money back on the first month. Email guru@graniteai.co and we refund it.

Start 7 days free

Mica Social

An AI that writes, designs, checks and publishes a post to your Facebook and Instagram every day. You send a photo when you have one. It does the rest.

Who it is for: You know the feed should be alive every day and it is not.
Price: $49/ month
What you get: A finished post every day, with a photo made for it. Nothing to shoot
Timed to your local weather, season and county numbers
Five checks before anything publishes. No approvals, no calendar, no logins
Every photo you send becomes its own extra post
The facts: An agency running a daily feed bills $500 to $1,500 a month. 3 days free with a card on file.
Our guarantee: 30 days money back on the first month. Email guru@graniteai.co and we refund it.

See Mica Social

GraniteAI builds and runs AI systems for small and midsized businesses, out of Manchester, New Hampshire.

Keep experimenting

One key. A catalog you can change.

Start with Poolside: Laguna S 2.1 (free). Return when the free catalog changes, choose a different model, and copy a fresh setup. The explicit :free id keeps the boundary visible.

omp docs Poolside: Laguna S 2.1 (free) on OpenRouter