What is Remy

Learn Remy’s main capabilities, ways to work, and operating boundaries, then choose a suitable first task.

1. Product overview and audience

1.1 What Remy is

Build it. Ship with confidence.

Remy is Corepass’s AI coding agent for individual developers and engineering teams. It combines project code, documents, and connected tools to take real tasks from requirements and implementation to checks and results you can review.

Describe your goal in natural language, add requirements and constraints, provide feedback as the work progresses, and inspect file changes and actual check results before deciding how to deliver the work.

1.2 What makes Remy useful

• From requirements to implementation. Understand the problem in project context, find code, edit files, and refine the work using check results. External documents and services require the relevant connection and authorization.
• Models from different providers. Choose an open-source or proprietary model and available reasoning settings, or use Dynamic to select a model for the task and move to a stronger one when needed.
• Clear check results. Run tests, type checks, or builds supported by the project. See what actually ran, the results, and unverified items when judging whether the work meets your requirements.
• Start locally and continue in the cloud. In versions and accounts with Cloud support, hand off a task and eligible context for continued execution, then follow and review it through available clients.

1.3 Who Remy is for

Remy primarily serves individual developers and engineering teams. It can also help with research, documents, and other knowledge work. For your first task, choose a familiar project that runs and a small change with an outcome you can check.

2. Ways to use Remy and this invitation

This invitation focuses on the macOS desktop app. Access to other clients, Cloud, and integrations depends on your invitation, installed version, and account availability.

For installation, see Downloads, the Desktop Guide, and the CLI Guide.

3. What you can do with Remy

3.1 Coding and code review

• Understand a codebase, locate relevant files, and trace business logic.
• Implement features, fix bugs, refactor code, and create or edit files.
• Run project commands and inspect tests, type checks, and build results.
• Review changes against requirements, with issue locations, triggers, and suggested fixes.
• Prepare change summaries and delivery notes, including PRs in supported repository workflows.

Committing, pushing, creating PRs, and merging follow your instructions, repository access, and authorization.

3.2 Check whether the work is complete

Include acceptance criteria in your task. Ask Remy to explain which checks actually ran, what passed or failed, and what was not run. If a check fails, continue in the same conversation to investigate, fix, and rerun it.

In versions that provide the Verification tool, you can explicitly ask it to check completion against the original requirements and acceptance criteria. The model may also call it as needed. Check the execution record for whether it ran and what it covered.

Code checks, human review, and interaction with the actual interface provide different evidence. Check the result against your requirements even after tests pass. Computer Use and more comprehensive automated verification depend on the version, environment, and permissions. Not every task runs the same checks, and these workflows are not prerequisites for your first task in this invitation.

3.3 Research, writing, and organizing information

Read and compare sources, organize grounded conclusions, extract meeting decisions and actions, and draft or revise reports and other documents. Connect and authorize external services before using their materials.

3.4 Browsers and computer applications

Browser Use and Computer Use can read information, perform actions, and check results in supported pages and applications. These capabilities currently support macOS only. Available actions depend on the version, app, browser, and granted permissions.

4. Models and task collaboration

4.1 Choose a model or use Dynamic

Remy supports open-source and proprietary models from different providers, including OpenAI and Anthropic, along with Kimi, DeepSeek, Qwen, GLM, and Grok. Available models, reasoning settings, and allowances depend on your plan and the options shown in the app.

• Choose a model: Select one before sending the task and adjust reasoning settings where offered.
• Dynamic: Let Remy choose a model for the task and move to a stronger one if the work gets stuck. The automatically selected model is not currently displayed.
• Switching during execution: Users cannot currently switch models manually while a task is running. Dynamic’s automatic escalation is separate from manual selection.

Your Remy account provides access to supported models without a separate model-provider account or your own API key. See Model selection and automatic routing.

4.2 Local tasks and Cloud

Local tasks use the project files and environment on your computer. Cloud tasks run remotely and do not stop simply because your computer disconnects. If a task is waiting for information or authorization, it still needs your response to continue.

When moving local work to Cloud, you can transfer supported conversation content and choose whether to include eligible uncommitted changes. Review the confirmation and file list before uploading. Local dependencies, running services, credentials, and sign-in state do not automatically become available in the cloud.

Cloud changes do not automatically sync back to local files. Review the cloud result, then bring the changes into your project through a PR or your usual Git workflow. See the Cloud Guide.

4.3 Continue context and delegate work

• Continue the original chat: Add requirements, answer questions, and refine results in the same conversation.
• Branch a conversation: Explore another approach from an existing discussion. Branching does not undo file changes.
• Import history: Copy local conversations from Claude Code, Codex, Cursor, or grok-build into Remy and continue there. Original records remain read-only. Tool results, reasoning, and attachments are not copied.
• Use subagents: Delegate smaller parts of a task and combine their results. You can also explicitly request that work be divided.

4.4 Project instructions and extensions

Use project instructions, Skills, plugins, and MCP services to adapt Remy to your workflow. The information an external service exposes and the actions it allows depend on the connected account and granted permissions. See Extensions and integrations.

5. Accounts, trials, and teams

5.1 Accounts and the Max trial

Use your Remy/Corepass account through the registration and sign-in options currently provided by the product. Check Pricing and the app for the model access and allowances included with Free, Pro, Max, and other plans.

An activation code in this invitation grants Max trial benefits; it is not a sign-in password. Use your inviter’s link or code, then use the same account across clients. Your inviter will communicate the trial end date. If you already have a paid plan or another active trial, contact your inviter first. This trial does not override valid benefits purchased separately.

5.2 Team

Team supports member invitations, roles and seats, centralized billing, and usage management.

• Roles determine team administration permissions; seats determine a member’s product benefits.
• Members use their seat’s included allowance. Additional team Usage credits and spending limits are managed separately and do not automatically pool included allowances.
• Unassigning a seat retains team membership. Removing a member revokes team access and releases the seat without deleting the personal account.
• Team administration permissions do not replace authorization for agent actions, the operating system, or external services.

See Accounts, teams, and usage.

6. Data, permissions, and feedback

6.1 How data is handled

To complete a task, Remy processes relevant requests, project content, and tool results. Corepass services or selected model providers may process some of that context. Running a tool locally does not mean that all processing stays on your computer.

Whether sessions contribute to model improvement, the associated user controls, retention, and team rules depend on your plan, settings, and applicable policies. See Data and permissions, the Trust Center, and the Privacy Policy.

6.2 Control the scope of work

Add the project folders needed for the task and describe the goal, limits, and acceptance criteria. You can ask for read-only analysis before deciding whether to edit files. Installing dependencies, committing, pushing, and acting in external services should stay within your authorization.

Actual access and execution also depend on conversation permissions, operating-system permissions, connection setup, and action approvals. Check the current project, account, and target environment before important actions.

6.3 Feedback and help

• App issues: Open ··· → Send feedback in the relevant chat. Include reproduction steps, expected and actual behavior, your version and operating system, and screenshots when helpful.
• Response quality: Use thumbs-up or thumbs-down below the response and explain what was useful, incorrect, or missing.
• Refine the task: Describe the observed behavior and desired result in the original chat.
• Cannot sign in or use in-app feedback: Contact support@corepass.com or your inviter.

6.4 Next steps

Complete your first small task with the Quick Start, then explore the guides, FAQ, and troubleshooting as needed.

Revision history

Version	Date	Updated by	Change summary
v0.2	2026-10-22	Link	Align positioning, models and Dynamic, local and cloud work, check results, access scope, and Max trial terms with current information
v0.1	2026-10-08	Link	Initial version