"Description": "Capability matrix for ABP Studio AI Agent modes, built-in tools, Studio automation tools, connected MCP tools, subagents, and hidden tool boundaries."
"Description": "Capability matrix for ABP Studio AI Agent modes, built-in tools, Studio automation tools, connected MCP tools, and subagents."
}
```
@ -87,9 +87,9 @@ The `ask_user` tool lets the agent request missing information during a run. Pen
## Shell And Migration Tools
Agent mode can run shell commands through `run_shell_command`. Shell execution is permission-gated and can be serialized with other workspace operations when required.
Agent mode can run shell commands through `run_shell_command`. Shell execution is permission-gated.
The `add_migration` tool adds Entity Framework Core migrations for configured migration projects. Build and migration execution share a gate so concurrent sessions do not run incompatible `dotnet` operations at the same time.
The `add_migration` tool adds Entity Framework Core migrations for configured migration projects.
## Studio Automation Tools
@ -114,18 +114,6 @@ Agent mode can use enabled ABP Studio automation tools.
Some solution-structure tools are not exposed because their information is already injected into the system prompt. This avoids redundant tool calls.
## Hidden Tools
The following implemented Studio tool categories are intentionally hidden from the AI Agent:
- Custom command listing and execution.
- Kubernetes chart and service listing.
- Embedded browser tools.
These tools may exist in Studio for other features, but they are not part of the AI Agent's callable tool surface.
## Subagents
The `spawn_subagent` tool delegates bounded research work to read-only specialist subagents.
@ -149,18 +137,3 @@ Configured MCP server tools can be added to Agent mode when:
MCP resources are visible in settings and can be opened for inspection, but the Agent-mode callable surface is composed from MCP tools. Plan and Ask modes do not receive MCP tools.
This capability lets the AI Agent consume tools from external MCP servers. It does not make ABP Studio AI Agent an MCP server for external clients.
## Tool Execution Gates
ABP Studio serializes selected operations to reduce parallel-session conflicts.
| Gate | Serialized operations |
| --- | --- |
| Build/migration | `dotnet_build` and migration execution. |
| Install libs | Client library installation. |
| Generate proxies | C# and Angular proxy generation. |
| Analysis | Studio package analysis. |
| Run task | Per task name. |
| Custom command | Per custom command name, for internal command execution paths. |
These gates do not prevent parallel conversations; they serialize operations that would conflict at the workspace or process level.
ABP Studio integrates AI Agent features into the Git and GitHub workflow. The integration uses selected files, staged or working-tree diffs, review notes, GitHub issue/PR context, and image attachments to build prompts for dedicated AI review or coding sessions.
AI Agent workflows define repeatable actions that Studio can run before or after an agent task. Workflows are associated with run profiles and are selected before a session starts.
@ -28,18 +30,9 @@ AI Agent workflows define repeatable actions that Studio can run before or after
The shared/personal setting is fixed when the workflow is created.
## Workflow Structure
A workflow contains:
- `Id`
- `Name`
- `Description`
- `IsShared`
- `BeforeTasks`
- `AfterTasks`
Workflow configuration includes the workflow name, description, sharing mode, before steps, and after steps.
Each workflow task stores its action type and action-specific targets such as modules, packages, application names, folders, containers, task names, proxy options, or migration project path.
@ -27,11 +27,17 @@ The agent is solution-aware. Its system context includes the current solution, m
The selected mode determines the agent's effective permission boundary.
| Mode | Purpose | Tool boundary |
| --- | --- | --- |
| Agent | Coding and execution | Can read and write files, run shell commands, add migrations, run enabled Studio tools, call connected MCP tools, and update active plan steps. |
| Plan | Read-only planning | Can inspect code and documentation, create or update plan files, and use read-only subagents. It cannot modify solution files or run shell/write tools. |
| Ask | Read-only question answering | Can inspect code, search/read ABP documentation, fetch relevant AI skills, and use read-only subagents. It cannot create plans or modify files. |
### Agent
Agent mode is used for coding and execution. It can read and write files, run shell commands, add migrations, run enabled Studio tools, call connected MCP tools, and update active plan steps.
### Plan
Plan mode is used for read-only implementation planning. It can inspect code and documentation, create or update plan files, and use read-only subagents. It cannot modify solution files or run shell/write tools.
### Ask
Ask mode is used for read-only question answering. It can inspect code, search/read ABP documentation, fetch relevant AI skills, and use read-only subagents. It cannot create plans or modify files.
Mode selection is per session. A session keeps its selected mode for subsequent prompts unless the user changes it before sending the next message.
This document describes coding practices for ABP Studio AI Agent. It focuses on how to shape agent runs so the agent receives the correct context, uses the correct tools, and verifies changes with the same ABP Studio automation available to the developer.
Volosoft Agent
2 months ago
GitHub
