Use your Abacus.AI ChatLLM credits for AI Butler

Use your Abacus.AI ChatLLM credits for AI Butler

Updated October 17, 2026 · Meowport Pro · About 6 minutes

A ChatLLM subscription from Abacus.AI comes with a monthly pool of credits, and the same credits pay for its RouteLLM API. Meowport's AI Butler can use that API like any other OpenAI-compatible service: one key, one address, one model name. This guide shows the setup, what the credits cover, and where Meowport treats this provider a little more carefully.

Who this is for

• You have Meowport Pro. AI Butler is a Pro feature; a custom model service only chooses which model answers.
• You pay for ChatLLM (the "AI super assistant" from Abacus.AI), alone or in a team, and want its credits to cover the Butler too.
• You are comfortable with each request's conversation and the pet context it needs being sent to Abacus.AI under its terms, the same as with any custom provider.

What the credits cover

• ChatLLM costs US$10 per user per month and includes 20,000 credits a month. The optional Pro tier adds another US$10 and raises the pool to 30,000. A discounted first month comes with proportionally fewer credits.
• The credits are one pool for everything you do in ChatLLM: chat, images, and the RouteLLM API. API calls are charged from the same pool, and how many credits a call costs depends on the price of the model it used.
• Subscription credits reset to the full amount on your renewal date and do not roll over. Credits you buy on top (US$10 for 10,000) do roll over.
• When the pool is empty you can buy more. Abacus.AI's pages disagree on whether the auto-routing model stays available after that, so plan on named models stopping until the next renewal or a top-up. ChatLLM warns you at 75 percent use.
• The usage log in ChatLLM can group spending by source, so you can see what the Butler consumed.

Step 1: Get your RouteLLM API key

1. Sign in to ChatLLM.
2. Click the RouteLLM API icon in the lower-left corner, or open the RouteLLM API page directly: https://abacus.ai/app/route-llm-apis.
3. Create or copy your API key. Abacus.AI does not document whether the key is shown only once or how to revoke it, so store it somewhere safe right away and treat it like a password: anyone who has it can spend your credits.

Teams: every member has their own key and spends from the team's pool; the team owner is invoiced.

Step 2: Connect it in Meowport

1. Open Meowport and go to Settings → AI Butler.
2. Choose Custom model API (marked Advanced).
3. Fill in the three fields:
   • Base URL: https://routellm.abacus.ai/v1
   • API Key: paste the key from Step 1.
   • Model name: route-llm lets Abacus.AI pick a model for each request. For a fixed model, use its id from the list, for example gpt-5.5, claude-sonnet-4-6 or gemini-2.5-flash.
4. Tap Fetch models to confirm the key works. RouteLLM lists the models your plan can use, with their pricing; pick one from the list instead of typing if you prefer.
5. Tap Save model settings.

Enterprise workspaces use their own address of the form https://<workspace>.abacus.ai/v1; ask your administrator.

How Meowport treats this provider

Meowport recognizes a handful of providers by their address and gives those models its full trust. RouteLLM is not one of them yet, so the Butler runs it in its cautious mode: it checks the model's answer against its own rules before acting, it does not let the model look up extra records on its own, and it may ask you more often or decline an unusual request. Everything else works the same way: the model list, action cards, undo from the chat, and the Before saving choice.

To read a clinic report or a lab photo, tap Test this model's image support once in the settings. RouteLLM accepts images on most of its models, but Meowport verifies it against the model you chose before it offers the attachment.

Step 3: Try it

Open the Butler and say something small, for example "Mochi weighed 4.2 kg this morning." The Butler drafts a weight record and, depending on your Before saving setting, saves it or asks first. Whatever you choose, it always asks before changing a record you already have, and it never deletes one. Anything it saves can be undone from the chat.

Costs and limits

• Every Butler request spends credits from your ChatLLM pool. A small, cheap model costs very little; a frontier model costs more. The model list in ChatLLM shows the relative prices.
• route-llm chooses the model by how complex the request looks. That keeps most short Butler requests cheap, but the cost of a given request is not fixed.
• Abacus.AI documents no request rate limit for the included credits. Context windows depend on the model you pick.
• When the pool is exhausted the Butler's requests fail until you buy credits or the month renews. Meowport shows the provider's message and writes nothing.

Privacy and what is sent

With a custom model service, each request sends your message, any images you attached, and the pet records the Butler needs for that request to the provider, under the provider's terms and privacy policy. Abacus.AI states that it does not use your data for training, that its model providers are bound the same way, and that its models run in the United States; it does not publish how long API logs are kept. Meowport's privacy policy describes the custom-provider path. Apple's on-device model stays available: switch back any time in the same settings. To stop using RouteLLM, clear the key in Meowport and remove it in ChatLLM.

Troubleshooting

• Authentication error. The key was pasted incompletely or no longer exists. Copy it again from the RouteLLM API page.
• The model is not found. Model ids must match exactly, including the provider prefix some open models carry. Tap Fetch models and pick from the list.
• Requests fail near the end of the month. The credit pool is probably empty; check the usage log in ChatLLM.
• Image attachments are not offered. Run Test this model's image support in the settings, or choose a model that lists image input.

Your responsibilities

• Check the provider's pricing and terms yourself; this guide reflects October 17, 2026 and cannot follow later changes.
• Keep the API key private. Do not paste it into chats, screenshots or shared notes.
• Watch your usage in ChatLLM. You, not Meowport, are billed for what the key spends.
• What you send is handled by the provider under its own policies, outside Meowport's control.
• AI can be wrong. Review what the Butler writes, and talk to your veterinarian before any medical decision.
• A Meowport is a personal record, not a certificate.