bootspring_assist | Bootspring Docs

Intelligent assistant that understands natural language requests and provides proactive suggestions.

Overview#

The bootspring_assist tool is the "magic" tool that analyzes natural language input and suggests the best actions. It detects intent, recommends agents, workflows, and specific tool calls based on your request.

Parameters#

Parameter	Type	Required	Description
`message`	string	Yes	The user's natural language request or question
`context`	object	No	Optional context about current work
`mode`	string	No	Mode: `suggest` (default), `explain`, or `execute`

Context Object#

Property	Type	Description
`currentFile`	string	Path to the file being worked on
`error`	string	Error message to debug
`recentActions`	array	List of recent actions taken

Mode Options#

Mode	Description
`suggest`	Returns suggestions and recommended tool calls (default)
`explain`	Provides reasoning behind suggestions
`execute`	Takes action directly

Usage Examples#

Basic Request#

Use the bootspring_assist tool with:
- message: "Help me set up authentication"

Response:

## Understanding Your Request

**Intent**: setup_auth
**Confidence**: 85% (high)

### Suggestion
I'll help you set up authentication for your project

### Recommended Action
Use `bootspring_agent` with:
```json
{
  "action": "invoke",
  "name": "security-expert",
  "topic": "Help me set up authentication"
}

Reason: High confidence (85%) match for "setup_auth"


### With Error Context

Use the bootspring_assist tool with:

message: "This is failing and I don't know why"
context: { "error": "PrismaClientKnownRequestError: Foreign key constraint failed" }


**Response:**
```markdown
## I Think I Understand

**Likely intent**: debug
**Confidence**: 72%

### Other Options
1. Analyze the error

### Recommended Action
Use `bootspring_agent` with:
```json
{
  "action": "invoke",
  "name": "database-expert",
  "topic": "Debug this error: PrismaClientKnownRequestError: Foreign key constraint failed"
}


### Low Confidence Request

Use the bootspring_assist tool with:

message: "help"


**Response:**
```markdown
## Need More Information

### Clarification Needed
I'm not sure what you need. Here are some things I can help with:

### Other Options
1. Show what I can do
2. See available experts
3. Check project status

Confidence Tiers#

The assist tool categorizes responses by confidence level:

Tier	Confidence	Behavior
High	80%+	Direct suggestion with specific tool call
Medium	50-80%	Multiple options offered
Low	30-50%	Asks for clarification, provides suggestions
Very Low	<30%	Shows capabilities and asks for more detail

Response Format#

Understanding Section#

For high confidence requests:

## Understanding Your Request

**Intent**: [detected intent]
**Confidence**: [percentage] (high)

For medium confidence:

## I Think I Understand

**Likely intent**: [detected intent]
**Confidence**: [percentage]

Suggestions Section#

### Suggestion
[Primary recommendation]

### Other Options
1. [Alternative 1]
2. [Alternative 2]
3. [Alternative 3]

Memory Insight#

When the system has learned from past actions:

### Learning Insight
[Insight from previous similar tasks]
Recommendation: [What worked before]

Recommended Tool Call#

### Recommended Action
Use `[tool_name]` with:
```json
{
  "param": "value"
}

Reason: [Why this tool was selected]


## Error Inference

When an error is provided in context, the assist tool automatically infers the best agent to consult:

| Error Contains | Suggested Agent |
|----------------|-----------------|
| prisma, sql, database | database-expert |
| auth, token, permission | security-expert |
| react, component, render | frontend-expert |
| api, fetch, request | backend-expert |
| test, jest, vitest | testing-expert |
| (other) | code-review-expert |

## Integration Example

### Workflow: Debug a Problem

Use bootspring_assist with:
- message: "My API endpoint is returning 500 errors"
- context: { "error": "TypeError: Cannot read property 'id' of undefined" }
Follow the recommended tool call to invoke the appropriate agent
Apply the agent's suggestions


## Best Practices

### Be Specific
Instead of: "Help me"
Try: "Help me add pagination to my user list API endpoint"

### Include Context
Provide relevant information:
- Current file being worked on
- Error messages
- What you've already tried

### Use Appropriate Mode
- `suggest` for exploring options
- `explain` when you want to understand why
- `execute` when you trust the AI to act

## Related Tools

- [bootspring_agent](/docs/mcp-tools/agent) - Invoke expert agents directly
- [bootspring_workflow](/docs/mcp-tools/workflow) - Orchestrate complex workflows
- [bootspring_skill](/docs/mcp-tools/skill) - Apply code patterns