AGENT USAGEPlanned

Make the next thing happen.

A product idea to try. A report to put together. A task you keep doing by hand. Agent is a planned AI usage card for assistants that work with code, files, and tools—toward a prototype, a document, or a reusable script.

For independent builders, solo founders, and small teams working with code and files.

Try an ideaProduce a deliverableReuse a workflow

Explore the planned uses

Start with a task. Work toward something you can inspect.

Give the work a concrete destination. These are the task directions planned for Agent.

Agent — Explore what you can do
What you want to do	What you can take forward	Where to start
Give an idea a working version	What you can take forwardEditable web prototypes, small tools, code changes, and run or test results.	Where to startA specific requirement, a project, and a clear definition of done.
Turn a pile of material into a deliverable	What you can take forwardStructured documents, information tables, report drafts, and organized files.	Where to startMaterial the agent may access and the file format you need at the end.
Make repeated work reusable	What you can take forwardScripts, processing steps, and tool workflows you can use for later tasks.	Where to startA recurring task, example inputs, and the tools it needs.

How the workflow is planned

Name the thing you want at the end
Provide the task, relevant material, and completion criteria: a working page, for example, or a structured report.
Connect the assistant and its tools
Choose a compatible agent app, configure its model connection, and grant access to the files and tools needed.
Review the result and keep going
Inspect file changes, run results, or report content. Refine the task before the next step.

Compatible software

Choose an app and open its setup guide. These instructions follow official software documentation; individual AddMana connections have not all been tested. Each card uses its own key and balance. Update the app configuration whenever you rotate its key.

Choose an editor, terminal, or workflow tool. The Agent card is not yet for sale; these are software setup instructions. Confirm models and tool permissions against the card when available.

Editor · Custom API

Cline

Work on code changes and tasks in a project folder.

Setup and usage guide · Cline

Before you start：The card needs a model suited to tool use; file and command permissions are configured in the app.

In Cline settings, select OpenAI Compatible as the API Provider.
Set Base URL to https://gateway.addmana.com/v1; enter this card’s key and exact model ID.
Set context and image capabilities to match the model. Start by reading a test file, then review one edit.

Check the result：Confirm the file was read correctly and the edit is reviewable. Approve command permissions as prompted.

Official software guide

Terminal · Custom model

pi

Work through code, files, and successive tasks from a terminal.

Setup and usage guide · pi

Before you start：Install pi; you need to edit JSON and set an environment variable.

Open ~/.pi/agent/models.json. Preserve existing entries and add addmana under providers.
Set baseUrl to https://gateway.addmana.com/v1, api to openai-completions, and authHeader to true.
Set apiKey to $ADDMANA_API_KEY and define that variable in the environment used to launch pi. Add the card’s exact model ID as id in the models array.
Enter /model in pi, select the new model, and ask it to read a test file.

Check the result：Confirm a reply and a file read. For authentication or model errors, check the variable, model ID, and card status.

Official software guide

Terminal · Custom API

Aider

Request changes in a code repository and review the resulting diff.

Setup and usage guide · Aider

Before you start：Install Aider and start in a test repository or branch.

Set OPENAI_API_BASE to https://gateway.addmana.com/v1 and OPENAI_API_KEY to this card’s key in the launch environment.
In your project, run aider --model openai/MODEL_ID, replacing MODEL_ID with the card’s exact model ID.
Add the files you want to work on, request a small change, and review the diff.

Check the result：Confirm only intended files changed and project checks pass. Follow Aider’s guidance for unknown-model warnings.

Official software guide

Workflow · Custom API

n8n

Connect forms, files, and model calls into an input-to-output workflow.

Setup and usage guide · n8n

Before you start：Set up n8n. Tool use needs model support; external service accounts are configured separately.

Create OpenAI credentials with this card’s key and Base URL https://gateway.addmana.com/v1.
Add AI Agent and OpenAI Chat Model. Select the credentials and card model. Start with Use Responses API off to use Chat Completions.
Connect a test tool and a manual trigger, then run with test data.

Check the result：Inspect the model reply, tool output, and node execution record. External tools follow their own billing.

Official software guide

Software licenses, web search, reference databases, and rendering may have separate charges. API connections, workflow configuration, and file imports are different ways to use these tools.

A few things to know before you choose

Should I choose Text or Agent?

Start with Text for reading, translation, and drafting. Consider the planned Agent card when an assistant needs to edit project files, run code, or call tools. Its actions depend on the app, installed tools, and permissions.

Does the card include an assistant that works on its own?

The card is planned to provide model usage inside a compatible agent app. The app, external tools, and their subscriptions are separate. Tested connection options will be listed at launch.

Can I start a project with it today?

Agent is in preparation and is not yet for sale. Models, prices, and supported workflows will be published when it launches.

Your next deliverable starts with a clear task.

Agent is in preparation. Compare the four cards to find the direction that fits your work.

Compare cards

Explore the other cards

Reading and writing. Tasks and tools. Visual creation. Research. Choose for the work ahead.

TextMake sense of what you read. Make something of what you think.MediaShape an idea into images, shots, and visual material for your next release.ResearchFind sources, compare perspectives, and keep evidence beside your conclusions.

Make the next thing happen.

For independent builders, solo founders, and small teams working with code and files.

Try an ideaProduce a deliverableReuse a workflow

Give an idea a working version

What you can take forwardEditable web prototypes, small tools, code changes, and run or test results.

Where to startA specific requirement, a project, and a clear definition of done.

Turn a pile of material into a deliverable

What you can take forwardStructured documents, information tables, report drafts, and organized files.

Where to startMaterial the agent may access and the file format you need at the end.

Make repeated work reusable

What you can take forwardScripts, processing steps, and tool workflows you can use for later tasks.

Where to startA recurring task, example inputs, and the tools it needs.

How the workflow is planned

Name the thing you want at the end

Provide the task, relevant material, and completion criteria: a working page, for example, or a structured report.

Connect the assistant and its tools

Choose a compatible agent app, configure its model connection, and grant access to the files and tools needed.

Review the result and keep going

Inspect file changes, run results, or report content. Refine the task before the next step.

Compatible software

Choose an editor, terminal, or workflow tool. The Agent card is not yet for sale; these are software setup instructions. Confirm models and tool permissions against the card when available.

Editor · Custom API

Cline

Work on code changes and tasks in a project folder.

Setup and usage guide · Cline

Before you start：The card needs a model suited to tool use; file and command permissions are configured in the app.

In Cline settings, select OpenAI Compatible as the API Provider.
Set Base URL to https://gateway.addmana.com/v1; enter this card’s key and exact model ID.
Set context and image capabilities to match the model. Start by reading a test file, then review one edit.

Check the result：Confirm the file was read correctly and the edit is reviewable. Approve command permissions as prompted.

Official software guide

Terminal · Custom model

pi

Work through code, files, and successive tasks from a terminal.

Setup and usage guide · pi

Before you start：Install pi; you need to edit JSON and set an environment variable.

Open ~/.pi/agent/models.json. Preserve existing entries and add addmana under providers.
Set baseUrl to https://gateway.addmana.com/v1, api to openai-completions, and authHeader to true.
Set apiKey to $ADDMANA_API_KEY and define that variable in the environment used to launch pi. Add the card’s exact model ID as id in the models array.
Enter /model in pi, select the new model, and ask it to read a test file.

Check the result：Confirm a reply and a file read. For authentication or model errors, check the variable, model ID, and card status.

Official software guide

Terminal · Custom API

Aider

Request changes in a code repository and review the resulting diff.

Setup and usage guide · Aider

Before you start：Install Aider and start in a test repository or branch.

Set OPENAI_API_BASE to https://gateway.addmana.com/v1 and OPENAI_API_KEY to this card’s key in the launch environment.
In your project, run aider --model openai/MODEL_ID, replacing MODEL_ID with the card’s exact model ID.
Add the files you want to work on, request a small change, and review the diff.

Check the result：Confirm only intended files changed and project checks pass. Follow Aider’s guidance for unknown-model warnings.

Official software guide

Workflow · Custom API

n8n

Connect forms, files, and model calls into an input-to-output workflow.

Setup and usage guide · n8n

Before you start：Set up n8n. Tool use needs model support; external service accounts are configured separately.

Create OpenAI credentials with this card’s key and Base URL https://gateway.addmana.com/v1.
Add AI Agent and OpenAI Chat Model. Select the credentials and card model. Start with Use Responses API off to use Chat Completions.
Connect a test tool and a manual trigger, then run with test data.

Check the result：Inspect the model reply, tool output, and node execution record. External tools follow their own billing.

Official software guide

Software licenses, web search, reference databases, and rendering may have separate charges. API connections, workflow configuration, and file imports are different ways to use these tools.

A few things to know before you choose

Should I choose Text or Agent?

Does the card include an assistant that works on its own?

The card is planned to provide model usage inside a compatible agent app. The app, external tools, and their subscriptions are separate. Tested connection options will be listed at launch.

Can I start a project with it today?

Agent is in preparation and is not yet for sale. Models, prices, and supported workflows will be published when it launches.