Agents
Agents are specialized AI expert profiles that provide deep domain knowledge. Instead of generic AI responses, agents give you guidance tailored to specific areas like frontend development, database design, or business strategy.
How Agents Work#
When you invoke an agent, Bootspring:
- Loads the agent profile with specialized instructions
- Combines with your project context from CLAUDE.md
- Provides enhanced guidance specific to that domain
Your Request
│
▼
┌─────────────────────────────────────┐
│ Bootspring Agent │
│ ┌─────────────┐ ┌──────────────┐ │
│ │ Agent │ │ Project │ │
│ │ Profile │ +│ Context │ │
│ │ │ │ │ │
│ └─────────────┘ └──────────────┘ │
└─────────────────────────────────────┘
│
▼
Expert Response
Agent Categories#
Technical Agents (Free)#
Core development expertise:
| Agent | Expertise | Use For |
|---|---|---|
frontend-expert | React, Next.js, CSS, accessibility | UI components, styling |
backend-expert | Node.js, APIs, server logic | Server implementation |
database-expert | SQL, Prisma, schema design | Data layer |
api-expert | REST, GraphQL, OpenAPI | API design |
testing-expert | Vitest, Jest, Playwright | Test writing |
security-expert | OWASP, auth, encryption | Security reviews |
performance-expert | Caching, optimization | Performance tuning |
devops-expert | CI/CD, Docker, Kubernetes | Infrastructure |
ui-ux-expert | Design systems, a11y | UX decisions |
architecture-expert | System design, patterns | High-level design |
code-review-expert | Best practices, quality | Code reviews |
vercel-expert | Vercel, Edge functions | Vercel deployment |
Platform Agents (Free)#
Platform-specific expertise:
| Agent | Expertise | Use For |
|---|---|---|
auth-expert | Clerk, NextAuth, Auth0 | Auth implementation |
payment-expert | Stripe, subscriptions | Payment systems |
email-expert | Resend, templates | Email features |
ai-integration-expert | LLMs, embeddings, RAG | AI features |
monitoring-expert | Sentry, logging, APM | Observability |
data-modeling-expert | ERD, normalization | Data design |
railway-expert | Railway deployment | Railway hosting |
mobile-expert | React Native, Flutter | Mobile apps |
infrastructure-expert | AWS, GCP, Azure | Cloud infrastructure |
Business Agents (Pro)#
Strategy and operations:
| Agent | Expertise | Use For |
|---|---|---|
business-strategy-expert | Strategy, GTM | Business planning |
growth-expert | PLG, retention, metrics | Growth tactics |
financial-expert | Modeling, projections | Financial planning |
fundraising-expert | Pitch decks, investors | Raising capital |
legal-expert | Contracts, compliance | Legal matters |
marketing-expert | SEO, content, campaigns | Marketing |
sales-expert | B2B sales, negotiations | Sales process |
product-expert | Roadmapping, OKRs | Product strategy |
operations-expert | Process, scaling | Operations |
competitive-analysis-expert | Market research | Competitive intel |
content-expert | Technical writing | Documentation |
research-expert | Tech evaluation | Research |
Enterprise Agents (Team)#
Enterprise-scale expertise:
| Agent | Expertise | Use For |
|---|---|---|
investor-relations-expert | Board, reporting | Investor comms |
partnerships-expert | Strategic BD | Partnerships |
private-equity-expert | M&A, due diligence | PE dealings |
Using Agents#
Natural Language#
Simply ask your AI assistant:
Use the frontend-expert agent to create a responsive data table.
Use the security-expert agent to review my authentication code.
Use the database-expert agent to optimize this slow query.
Best Practices for Requests#
Be Specific:
# Less effective
Use frontend-expert to help with the form.
# More effective
Use frontend-expert to create a multi-step checkout form with
React Hook Form, Zod validation, shipping/billing steps, and
proper error handling.
Provide Context:
Use database-expert to design a user schema. Context:
- Multi-tenant SaaS with organizations
- Need to support SSO via SAML
- Users can belong to multiple organizations
- Soft deletes required for compliance
Request Specific Output:
Use security-expert to review login.ts with outputFormat="review"
to get a structured security assessment.
Agent Chaining#
For complex features, chain multiple agents:
Feature Development Chain#
1. Architecture Phase
→ architecture-expert: Plan the overall approach
2. Data Layer
→ database-expert: Design the schema
→ api-expert: Design the API surface
3. Implementation
→ backend-expert: Build the API
→ frontend-expert: Build the UI
4. Quality
→ testing-expert: Write tests
→ security-expert: Security review
→ code-review-expert: Code quality
Example: Building Auth#
Step 1: Use architecture-expert to plan the auth system
→ Get high-level approach, component breakdown
Step 2: Use database-expert to design user/session schema
→ Get Prisma schema, indexes, relations
Step 3: Use auth-expert to implement Clerk integration
→ Get middleware, protected routes, hooks
Step 4: Use frontend-expert to build login/signup UI
→ Get components, forms, error handling
Step 5: Use testing-expert to write auth tests
→ Get unit and E2E tests
Step 6: Use security-expert to review implementation
→ Get security assessment, fixes
Agent Collaboration#
Some tasks benefit from multiple agent perspectives:
Security Audit#
Use security-expert, backend-expert, and database-expert together
to perform a comprehensive security audit of the payment system.
Performance Optimization#
Use performance-expert, frontend-expert, and database-expert
to optimize the dashboard load time.
Customizing Agents#
Custom Instructions#
Add project-specific behavior in bootspring.config.js:
1module.exports = {
2 agents: {
3 customInstructions: {
4 'frontend-expert': `
5 - Always use TypeScript strict mode
6 - Prefer server components in Next.js
7 - Use Tailwind CSS, never inline styles
8 - Include proper ARIA attributes
9 - Add loading and error states
10 - Use our Button, Card, Input components from @/components/ui
11 `,
12 'database-expert': `
13 - Use Prisma as the ORM
14 - Include createdAt/updatedAt on all models
15 - Use soft deletes (deletedAt field)
16 - Add indexes for foreign keys
17 - Use UUID for IDs, not auto-increment
18 `,
19 'testing-expert': `
20 - Use Vitest for unit tests
21 - Use Playwright for E2E
22 - Mock external services
23 - Minimum 80% coverage
24 - Use data-testid for E2E selectors
25 `,
26 },
27 },
28};Enabling/Disabling Agents#
1module.exports = {
2 agents: {
3 // Only allow specific agents
4 enabled: [
5 'frontend-expert',
6 'backend-expert',
7 'database-expert',
8 'testing-expert',
9 ],
10
11 // Or disable specific agents
12 // disabled: ['mobile-expert', 'devops-expert'],
13 },
14};Agent Selection Guide#
| Task | Primary Agent | Supporting |
|---|---|---|
| Build UI component | frontend-expert | ui-ux-expert |
| Create API endpoint | backend-expert | api-expert |
| Design database | database-expert | data-modeling-expert |
| Add authentication | auth-expert | security-expert |
| Optimize performance | performance-expert | frontend/backend |
| Set up CI/CD | devops-expert | - |
| Security review | security-expert | code-review-expert |
| Write tests | testing-expert | - |
| System architecture | architecture-expert | - |
| Deploy to Vercel | vercel-expert | devops-expert |
| Add Stripe payments | payment-expert | backend-expert |
| Business planning | business-strategy-expert | - |
| Raise funding | fundraising-expert | financial-expert |
Agent Output Formats#
Agents can provide different output types:
Guidance (default)#
Explanations, recommendations, and code examples.
Code#
Primarily code with inline comments.
Review#
Structured assessment with findings and recommendations.
Plan#
Step-by-step implementation plan.
Use security-expert with outputFormat="review" to assess login.ts
Tier Access#
| Tier | Agents Available |
|---|---|
| Free | Technical (12) + Platform (9) |
| Pro | Free + Business (12) |
| Team | Pro + Enterprise (3) + Custom |
| Enterprise | All + Unlimited Custom |
Troubleshooting#
Agent Not Found#
Check the exact agent name:
bootspring agent listGeneric Responses#
- Regenerate context:
bootspring generate - Be more specific in your request
- Add custom instructions in config
Need Unlisted Expertise#
Use the closest agent with custom context:
Use backend-expert with focus on WebSocket implementation
for real-time features.
Related#
- Using Agents - Getting started guide
- bootspring_agent - Tool reference
- Agent Collaboration - Advanced patterns
- Agents Reference - Full agent documentation