This page explains the main terms you see in Bold Create and how they fit together.
How the pieces fit together
- An integration connects Bold Create to your Git provider.
- Through it, you pick a repository and branch, or a saved environment.
- You choose an agent and a model, and optionally add skills.
- You send a prompt. This starts a thread and its first run.
- The agent works in a cloud environment, and you review the changes in the Git panel.
- You ask the agent to open a pull request.
Concepts
Thread
A thread is one task and its full history: your prompts, the agent’s replies, the steps it took, and the changes it made. Every thread is listed in the sidebar under Tasks. You can search threads with Search tasks and group them with Group tasks.
Send follow-up messages in the same thread to refine the work. The agent keeps the context of the thread, including its working branch.
Run
A run is one turn of the agent’s work, from the moment you send a message until the agent finishes, stops with an error, or you stop it. A thread can have many runs. The sidebar shows the status of the latest run:
| Status | Meaning |
|---|---|
| Running | The agent is working. |
| Completed | The agent finished. |
| Error | The run stopped because of a problem. Open the thread to see the message. |
Agent
The agent is the AI that does the work. Pick one in the agent picker below the prompt box. Bold Create has three built-in agents:
| Agent | Description in the product | Changes files |
|---|---|---|
| Agent | General-purpose AI agent for software engineering tasks | Yes |
| Ask | Answer questions about your codebase without making any changes | No |
| Plan | Produce a structured implementation plan without writing any code | No |
Besides the built-in agents, the agent picker can list repository agents defined in the selected repository. You can switch agents between messages in the same thread. Learn more on the Cloud and repository agents page.
Model
The model is the large language model the agent uses. Pick one in the model picker. Models are grouped into Premium Models and BYOK Models (models your organization connects with its own provider keys). Some models also let you set a thinking effort, which controls how much the model reasons before it acts. Higher effort improves accuracy on complex tasks, such as multi-file changes or hard bugs, but uses more tokens and so more credits. Lower effort uses fewer tokens and suits simple, well-scoped tasks. Learn more on the Models page.
Integration
An integration connects Bold Create to a Git provider so the agent can read your repositories and push its changes. Bold Create supports GitHub, GitLab, and self-hosted Gitea. Manage them on the Integrations page. Learn more on the Integrations page.
Repository and branch
The repository is the codebase the agent works on. The branch you pick is the starting point. The agent does not commit to that branch. It creates its own working branch, whose name starts with bc/, and uses it for every run in the thread.
Environment
An environment is a saved, reusable setup for the agent’s cloud workspace. It can include one or more repositories, setup and start commands, and environment variables. Use an environment when your project needs extra tools or several repositories to work. Share environments with your team or organization. Learn more on the Environments page.
Skill
A skill is a reusable set of instructions, such as a review checklist or coding conventions. Each skill is a SKILL.md file in its own folder. Skills come from your organization or from the repository, for example under .boldcreate/skills/<name>/. To use one, choose it from the Skills list in the prompt box menu (or type /) and describe your request. The agent can also find skills in the workspace and read the ones that match your request. Learn more on the Skills overview page.
Workspace panel: Background, Changes, Files, and Terminal
Open the workspace panel with the panel button in the top-right corner of the thread. It has four tabs:
- Background lists commands the agent left running (such as a dev server) and its sub-agents, with their status. You can stop a running one.
- Changes lists the files the agent changed, per repository, with a diff for each file. Add a review comment on a line to send feedback to the agent. Learn more on the Review changes page.
- Files is a file explorer and editor for the workspace.
- Terminal gives you shell access to the same workspace. You can open several terminals.
Files and Terminal appear only when the workspace supports them for the thread.
Pull request
When you ask, the agent commits and pushes its branch and opens a pull request in your Git provider. It writes the title and description. A pull request card with a link appears in the thread. Learn more on the Pull requests page.
Credits and usage
Runs use credits from your organization. Check your usage on the Dashboard, which you open from the profile menu.