Your code stays simple.
The agents don't.

PromptShuttle is a drop-in OpenAI replacement that orchestrates multi-agent workflows server-side. One API call in, final answer out — your app never touches agent logic.

// Just a normal OpenAI call
const response = await openai.chat.completions.create({
  model: "research-agent",
  messages: [
    { role: "user", content: "Analyze this document" }
  ],
});

Behind this call, PromptShuttle spawns sub-agents, routes across providers, runs tool calls, and returns the finished result.

An API, Not a Framework

Every competitor is an SDK — you embed agent logic in your application code. PromptShuttle is different: your app sends a standard API call, and orchestration happens server-side.

SDK Approach

Agent logic lives in your code
Change agents = redeploy your app
You manage providers, retries, costs
Each team builds their own plumbing

PromptShuttle

Agent logic lives in the proxy
Change agents = update config
Proxy handles providers, retries, costs
One deployment serves all teams

Built for Teams That Ship

Platform Teams

Offer AI capabilities to internal teams without each team learning an agent framework.

Learn more

Legacy & Existing Software

Make any software agent-capable with webhooks — no rewrite, no SDK.

Learn more

Agencies

Manage AI for multiple clients with built-in multi-tenant isolation.

Background Jobs

Use AI as a callable service from Trigger.dev, Temporal, or any job runner.

Everything You Need to Ship Agents

Multi-Agent Orchestration

Define agent flows in the dashboard. Sub-agents spawn, run tool calls, and aggregate results automatically.

OpenAI-Compatible API

Standard chat completions endpoint. Use the OpenAI SDK you already have — just change the base URL.

Multi-Provider Routing

Route to the best model for each agent. OpenAI, Anthropic, Google, DeepSeek — switch without code changes.

Cost Tracking & Budgets

See costs aggregated across the entire agent tree. Set budgets per tenant, per flow.

Multi-Tenant

One deployment serves all your teams or clients. Isolated keys, billing, and flow configurations.

Agent Observability

Built-in agent tree visualization, event streaming, and per-step cost breakdown.

See It in Action

A full dashboard for orchestration, observability, and cost tracking — not just an API.

Visualize Agent Orchestration

See the full execution graph for every request. Root agents, sub-agent spawns, tool calls, timing, and costs — all in one view.

Agent execution graph showing a root agent spawning three research sub-agents

Deep Flow Analytics

Track every flow's performance over time. Invocation trends, per-model breakdowns, and percentile latencies — all built in.

Flow analysis dashboard with invocation charts and per-model performance breakdown

Performance at a Glance

Runtime, cost, and tool call distributions help you spot outliers and optimize agent configurations before they become problems.

Runtime, cost, and tool calls distribution charts with model statistics

Usage & Cost Tracking

See costs aggregated by model and by flow. Know exactly where every dollar goes across your entire agent fleet.

Usage overview showing cost by model donut chart and cost by flow breakdown

First-Class MCP Support

Connect PromptShuttle to any MCP-compatible client — Claude Desktop, Cursor, Windsurf, Claude Code — and manage your agent infrastructure without leaving your editor.

Your AI manages your AI

PromptShuttle exposes a full MCP server so your coding assistant can create flows, update prompts, check analytics, and deploy changes — all through natural language.

Create and update agent flows
Edit prompts and tool configurations
Query usage analytics and costs
Manage environments and deployments
Works with any MCP-compatible client

// claude_desktop_config.json
{
  "mcpServers": {
    "promptshuttle": {
      "url": "https://app.promptshuttle.com/mcp"
    }
  }
}

One URL. No install, no CLI, no SDK.

Customer References

TenderStrike | AI tender analysis for the construction industry

TenderStrike uses PromptShuttle to proxy requests to a variety of LLMs, and to manage automated fallbacks and retries. Their team uses PromptShuttle as a collaboration tool between domain experts, business users (for reporting) and developers.

BRAND KARMA

BrandKarma | Generative Engine Optimization & Brand Insights

BrandKarma is an AI and LLM search insights platform for marketing teams and agencies. It monitors LLM presence of brands, helps users to find the best-performing content, identify gaps in their content strategy, and optimize their marketing and PR strategies for AI visibility.

Optivise | Automatic Product Data Optimization for E-commerce

Optivise uses PromptShuttle to enhance their AI-driven product data optimization platform. By leveraging multiple LLMs through PromptShuttle, they ensure high-quality, accurate, and engaging product descriptions that boost e-commerce performance across marketplaces such as ADEO, Debenhams, B&Q and more.

FacetFlux | AI product data enrichment for industrial catalogs

FacetFlux uses PromptShuttle to enrich, classify, and translate product data at catalog scale. Schema inference, ETIM/eCl@ss mapping, and bulk generation all route through the proxy so the product stays one API call away from any model.

veft

Veft | AI-native CRM for retained executive search

Veft uses PromptShuttle to run the language-model work behind candidate matching, longlist ranking, and client-ready documents. Search consultants stay in the product; routing, retries, and cost tracking sit in the proxy.

revuo

Revuo | The B2B SaaS directory LLMs cite

Revuo uses PromptShuttle to extract capabilities, verify listings, and answer agent queries. Multiple models run behind one API so the directory stays structured, current, and callable.

Straightforward Pricing

Start free, scale as you grow. Simple usage-based pricing with no agent orchestration surcharges.

Start Orchestrating in Minutes

Point your OpenAI SDK at PromptShuttle, define your first agent flow, and ship.

Your code stays simple. The agents don't.