Skip to Content

FAQ

Frequently asked questions about the isA platform.

General

What is isA?

isA is a complete platform for building, deploying, and scaling AI agents. It includes:

  • Agent SDK - Python library for building agents
  • MCP - 190+ pre-built tools for agents
  • Model Service - Unified LLM gateway
  • User Services - 35 microservices for auth, payments, storage
  • Cloud - Production-ready Kubernetes infrastructure

Is isA open source?

Yes, isA is open source. You can find all repositories at github.com/xenoISA .

What models does isA support?

isA supports multiple LLM providers:

ProviderModels
AnthropicClaude claude-sonnet-4-20250514, Claude Opus, Claude Haiku
OpenAIGPT-4o, GPT-4o-mini, GPT-4 Turbo
GoogleGemini Pro, Gemini Ultra
LocalAny Ollama-compatible model

How much does isA cost?

  • Self-hosted: Free (you pay for infrastructure)
  • Cloud hosted: Usage-based pricing (planned — see Platform for updates)

Model costs are passed through at provider rates.

Agent SDK

How do I install the SDK?

pip install isa-agent-sdk

How do I create a simple agent?

The fastest way is with query() or ask():

from isa_agent_sdk import query, ask
 
# Streaming (recommended)
async for msg in query("What is 2+2?"):
    if msg.is_text:
        print(msg.content, end="")
 
# One-shot (returns final text)
response = await ask("What is 2+2?")
print(response)

Or use the Agent class for reusable agents:

from isa_agent_sdk import Agent, ISAAgentOptions
 
agent = Agent(
    name="my-agent",
    options=ISAAgentOptions(
        allowed_tools=["web_search", "calculator"],
        model="claude-sonnet-4-20250514",
    )
)
 
result = await agent.run("What is 2+2?")
print(result.text)

Can agents remember previous conversations?

Yes, use session_id to maintain context across calls:

from isa_agent_sdk import query
 
async for msg in query("My name is Alice", session_id="session-123"):
    print(msg.content, end="")
 
# Later, same session resumes context
async for msg in query("What's my name?", session_id="session-123"):
    print(msg.content, end="")  # "Your name is Alice"

See Memory & State for persistent memory backends (Redis, PostgreSQL, Qdrant).

How do I add custom tools?

from isa_agent_sdk import Agent, ISAAgentOptions, tool
 
@tool
def my_custom_tool(query: str) -> str:
    """Search my database."""
    return database.search(query)
 
agent = Agent(
    name="custom-agent",
    options=ISAAgentOptions(tools=[my_custom_tool])
)

Can I use multiple models in one agent?

Yes, specify the model in options or per-call:

from isa_agent_sdk import query
 
# Use a specific model
async for msg in query("Complex reasoning task", model="claude-opus-4-20250514"):
    print(msg.content, end="")
 
# Or configure in options
from isa_agent_sdk import ISAAgentOptions
options = ISAAgentOptions(model="gpt-4o-mini")
async for msg in query("Quick task", options=options):
    print(msg.content, end="")

MCP Tools

What tools are available?

Over 190 tools across categories:

  • Search: web_search, image_search, news_search
  • Files: file_read, file_write, file_list, file_delete
  • Code: code_interpreter, shell, git
  • Browser: browser_navigate, browser_click, browser_screenshot
  • Database: postgres_query, redis_get, neo4j_query
  • APIs: http_request, graphql_query

See the full tools list.

How do I use a tool directly?

from isa_mcp import tools
 
result = await tools.web_search("latest AI news")

Can I create custom MCP tools?

Yes, see the tool development guide.

Deployment

What are the system requirements?

Minimum (development):

  • 4 CPU cores
  • 8 GB RAM
  • 50 GB disk

Recommended (production):

  • 8+ CPU cores
  • 32+ GB RAM
  • 200+ GB SSD
  • Kubernetes cluster

How do I deploy locally?

cd deployments/kubernetes/local/scripts
./kind-setup.sh
./kind-deploy.sh

How do I deploy to production?

  1. Set up a Kubernetes cluster (EKS, GKE, or self-managed)
  2. Install ArgoCD
  3. Connect your repository
  4. Push to main branch

See the deployment guide.

Can I use my own infrastructure?

Yes, isA is fully self-hostable. You can:

  • Use your own Kubernetes cluster
  • Bring your own databases
  • Use your own model API keys

Security

How is data encrypted?

  • At rest: AES-256 encryption
  • In transit: TLS 1.3
  • Secrets: Stored in Vault or AWS Secrets Manager

How does authentication work?

isA uses JWT tokens for API authentication:

  1. User authenticates via /auth/login
  2. Receives JWT access token + refresh token
  3. Include token in Authorization: Bearer <token> header
  4. Token validated by APISIX gateway

Is there RBAC support?

Yes, isA has fine-grained role-based access control:

  • Organizations with multiple roles
  • Resource-level permissions
  • API scope restrictions

See authentication docs.

Is isA SOC 2 compliant?

Self-hosted deployments inherit your compliance posture. Cloud-hosted SOC 2 Type II compliance is planned.

Troubleshooting

Agent not responding?

  1. Check your API key is set: echo $ISA_API_KEY
  2. Verify model service is running
  3. Check logs: kubectl logs -l app=model-service

Tools not working?

  1. Verify tool is installed: isa tools list
  2. Check MCP service: curl http://localhost:8081/health
  3. Review tool permissions

Slow responses?

  1. Enable caching in model service
  2. Use streaming for long responses
  3. Consider a faster model (e.g., gpt-4o-mini)

Memory issues?

  1. Clear old memories: agent.memory.clear()
  2. Use memory limits: memory={"max_items": 100}
  3. Check Redis memory: redis-cli INFO memory

Getting Help

Where can I get support?

How do I report a bug?

  1. Search existing issues 
  2. Create a new issue with:
    • isA version
    • Steps to reproduce
    • Expected vs actual behavior
    • Logs if available

How do I request a feature?

Open a GitHub Discussion  with:

  • Use case description
  • Proposed solution
  • Alternatives considered
Home