Hrida AI Blog

How to Use Spro as a Super Intelligence Gateway for Every LLM

Sign up, generate your API key, and run any OpenAI, Anthropic, or Google model through the SPRO Super Intelligence Gateway with intelligent routing, real-time alerts, and built-in security guardrails, powered by Hrida AI.

← Back to Blog

Get Started with SPRO: Your Super Intelligence Gateway for Any LLM

Go from sign-up to your first model response in minutes. Connect to multiple AI providers through a single gateway, with intelligent routing, cost controls, real-time alerts, and security guardrails built into the workflow.

As teams increasingly adopt multiple LLM providers, managing AI infrastructure becomes more complex. Different providers mean separate SDKs, API keys, integrations, monitoring dashboards, and security configurations. This fragmented approach increases operational overhead, makes costs harder to track, and creates more points of failure across your AI stack.

SPRO by Hrida AI addresses this challenge by introducing a unified intelligence layer between your applications and AI models. Instead of integrating each provider independently, your application communicates with SPRO through a single API gateway. SPRO handles model routing, forwards requests to the appropriate provider, and brings responses back through a centralized control layer where routing policies, security guardrails, and monitoring can be applied.

Whether you're working with OpenAI, Anthropic, Google, or other supported LLM providers, SPRO simplifies how you connect, manage, secure, and monitor your AI workloads.

In this guide, we'll walk through the complete setup process, from creating your SPRO account and generating your API key to making your first model request and configuring guardrails for production workloads.

No additional infrastructure to install. No need to maintain separate gateway integrations for every provider. Just one unified gateway to access and manage your AI ecosystem.

1. Sign up

Step 1: Create Your SPRO Account

Get started by visiting spro.hridaai.com and creating your account.

Once you sign up, SPRO automatically provisions a default workspace, giving you an environment to start configuring your AI gateway and making your first model request. No additional setup or infrastructure configuration is required to get started.

Understanding Workspaces

A workspace acts as a centralized environment for managing your AI integrations and configurations. It brings together:

  • API Keys: Create and manage the credentials your applications use to access SPRO.
  • Guardrails: Configure security, privacy, and policy enforcement controls for your AI requests.
  • BYOK Connections: Connect your existing OpenAI, Anthropic, Google, and other supported provider credentials.
  • Observability: Monitor API activity, token consumption, costs, request traces, and model performance.

As your usage expands, you can create additional workspaces to isolate different teams, applications, or environments. For example, you can maintain separate workspaces for Development, Staging, and Production to keep API keys, security policies, provider connections, and observability data independently managed.

2. Create an API Key

Once your workspace is ready, navigate to the API Keys section and generate a new API key.

SPRO API keys use the spro- prefix and authenticate requests through the standard Bearer Token authorization mechanism. This allows you to connect your applications to the SPRO gateway without managing separate gateway credentials for each AI provider.

Your API key acts as the authentication credential for accessing SPRO, so treat it as a sensitive production secret.

Security best practices:

  • Use environment variables: Store your API key in environment variables or a dedicated secrets manager.
  • Keep credentials server-side: Never expose your API key in frontend or browser-side JavaScript.
  • Avoid public exposure: Do not commit API keys to Git repositories, publish them in documentation, or include them in publicly accessible code examples.
  • Restrict access: Share credentials only with authorized team members and services that require them.
  • Rotate compromised keys: If a key is accidentally exposed, revoke it and generate a replacement immediately.

Once generated, copy your API key and securely configure it in your application's environment. You'll use this key in the authorization header when making requests to SPRO.

Example Authorization Header:

Authorization: Bearer spro-your-api-key

3. Point Your Client to SPRO

SPRO is OpenAI-compatible, so connecting your application requires just a small configuration change. Set the base URL to https://spro.hridaai.com/v1 and authenticate using your SPRO API key.

Example using Python:

from openai import OpenAI

client = OpenAI(
    base_url="https://spro.hridaai.com/v1",
    api_key="SPRO_API_KEY"
)

completion = client.chat.completions.create(
    model="claude-fable-5.2",
    messages=[
        {"role": "user", "content": "Summarize this ticket in two sentences."}
    ]
)

print(completion.choices[0].message.content)

That's it! Your request now runs through SPRO.

The same API key also works with the OpenAI Responses API (/v1/responses) and Anthropic Messages API (/v1/messages). You can also connect coding assistants like Cursor, Claude Code, Cline, and Aider through SPRO.

4. Choose any model

Browse the model catalog, pick a model and put its ID in the model field. Availability changes over time, so confirm the exact ID before you ship. You can control where a request goes at three levels:

StyleExampleWhat it does
Bare nameglm-5.3-flashSpro routes the request automatically
Author prefixz-ai/glm-5.3-flashPins the model's author
Host prefixtogether/glm-5.3-flashPins the provider that hosts the model

Switching models is a one-line change, not a new integration. Models do differ in the parameters they accept. Some do not support temperature or top_p, for example, so check each model's page.

5. Test the Features Your Application Relies On

Before moving to production, test SPRO with inputs that reflect your actual application workloads. A model that performs well in a demo may behave differently under real-world conditions.

Focus on the capabilities your application uses:

  • Extraction: Verify structured outputs, required fields, and JSON responses.
  • Chat: Test response quality and streaming behavior.
  • Tool Calling: Validate function calls using realistic tool definitions.
  • Coding: Run generated code against your existing test cases.
  • Reasoning: Compare reasoning levels across quality, latency, and cost.

Monitor response headers to understand request behavior:

HeaderPurpose
x-spro-resolved-modelIdentifies the model that served the request.
x-spro-attemptsShows the number of attempts made.
x-spro-fallback-fromIdentifies the original model when a fallback is used.
x-spro-cacheShows the cache outcome.

For multi-step agent workflows, use the x-spro-trace-id header to group related API calls under a single trace. You can also retrieve past request details using the request ID, including provider, model, token usage, cost, and execution time.

6. Control Cost, Speed, and Reliability

SPRO gives you control over how requests are routed, how quickly they are processed, and how much they cost.

Service Tiers

Choose a service tier based on your application's performance and budget requirements. You can specify it using a model ID suffix or the service_tier parameter. If neither is specified, SPRO defaults to Standard.

TierSuffixBest for
StandardNoneRegular workloads at standard pricing
Flex:flexLower-cost requests that can tolerate queuing
Priority:priorityWorkloads requiring lower latency
Ultrafast:ultrafastApplications requiring the lowest latency

Routing and Failover

SPRO uses failover routing by default. If the primary provider becomes unavailable or returns an error, SPRO attempts other providers serving the same model in priority order.

You can also configure priority routing, pin a specific provider, or define your own ordered fallback model chain.

Choose fallback models carefully. Switching providers for the same model and switching to an entirely different model can produce different results. Test both response quality and execution time before deploying your routing strategy.

Limits and Caching

Set request, token, and cost limits to prevent unexpected usage from consuming your balance, especially in automated workflows and AI agent loops.

SPRO also supports:

  • Exact-match caching: Reuses responses for identical requests.
  • Semantic caching: Reuses responses for requests with similar meanings.

Cache entries have a configurable time-to-live (TTL), defaulting to 600 seconds.

Alerts

SPRO provides email and in-app alerts to help you monitor usage, costs, and operational changes.

Available alerts include:

  • Error Rate Spike: Notifies you when request failures exceed your configured threshold.
  • Low Balance & Balance Depletion: Warns you when your available balance is running low.
  • Dormant Key: Identifies API keys that have not been used for a specified period.
  • Model Usage Report & Dashboard Snapshot: Provides scheduled usage and spending summaries.
  • Model Deprecation & Price Change: Notifies you about model retirement or pricing updates.
  • New Model Release: Keeps you informed when new models become available.

New workspaces have Low Balance alerts enabled by default. You can configure the remaining alerts through your workspace settings.

7. Add Security Guardrails

SPRO guardrails inspect both incoming prompts and outgoing model responses, helping enforce security and content policies across your AI applications.

You can configure guardrails for:

  • Sensitive Information Detection: Detects and masks PII, PHI, PFI, and API keys before requests reach providers.
  • Prompt Injection Protection: Detects prompt injection, jailbreak attempts, hidden Unicode characters, and obfuscated attacks.
  • Content Moderation: Filters harmful, toxic, or abusive content.
  • Output Validation: Supports JSON structure validation, hallucination detection against supplied context, and refusal detection.

You can configure the appropriate action for each guardrail based on your security requirements.

For additional privacy, enable Zero Data Retention (ZDR) at the workspace level. This prevents prompt and response content from being stored while retaining operational metadata such as token counts, latency, and status codes for billing and monitoring.

You can also connect your existing provider accounts using Bring Your Own Key (BYOK) while continuing to use SPRO's routing, guardrails, and observability features.

Guardrails add an important layer of protection, but they do not guarantee complete prevention of data exposure or attacks. Test configurations against your application's workloads and combine them with your broader security controls.

8. Roll Out to Production

Before switching your entire application to SPRO, start with a gradual rollout.

Compare model response quality, error rates, latency, and overall costs against your existing setup. Test normal workloads, rate limits, and provider failures, and maintain a rollback path until you're confident in the results.

Once everything is validated, gradually expand traffic and onboard additional models as your application requirements evolve.

Ready to get started? Sign up for SPRO, generate your API key, and send your first request. One gateway is all you need to start connecting and managing your AI models.

Hrida AI
Tue Nov 17 2026