Interference
NewOpenAI-compatible Gateway with Instant Indian Rupee (₹) Settlement is live
Get API Key →
[GATEWAY V1.0.0]

The open inference gateway for AI agents

Interference brings agentic inference to developers worldwide. Offering generous limits and reliable access to the most capable open-source models with transparent streaming tool-calling, atomic two-phase credit holds, and localized Rupee (₹) billing.

⚡ CLAUDE SONNET 5⚡ GPT-5.4⚡ GEMINI 3.7 FLASH⚡ DEEPSEEK V4 FLASH

Use with any agent. ₹1 = 100 Credits. Top up balance when needed via UPI & Cards. Cancel or withdraw any time.

curl -fsSL https://api.interference.dev/v1/models -H "Authorization: Bearer sk_live_..."

|

Gatewayclaude-sonnet-5
STATUS: 200 OK
tab switch modelctrl+p gateway commands

What is Interference?

Interference is an open inference gateway that allows developers to run cutting-edge open models with zero SDK changes, low routing latency, and mathematical balance guarantees.

  • [*]
    Zero SDK changesDrop-in OpenAI SDK compatibility. Simply override the baseURL to point to Interference.
  • [*]
    Transparent tool callingStreams multi-turn tool parameters without executing code on foreign servers. Local agent sovereignty.
  • [*]
    Two-phase atomic credit ledgerPre-flight reservations lock tokens before inference and settle afterward. Zero negative balance drift.
  • [*]
    D1 Authoritative StorageACID transactional double-entry ledger paired with low-latency edge KV authentication caches.
  • [*]
    Rupee (₹) native settlementDirect UPI & Indian Netbanking top-ups via Dodo Payments. No foreign exchange currency fees.
  • [*]
    Any agent frameworkCompatible with Vercel AI SDK, LangChain, Cursor, OpenCode, AutoGen, and OpenAI Python.
Read docs

How Interference works

Integrate in 60 seconds with 3 simple steps.

[1]
Create an account- sign in to the console and generate your secret API key.
[2]
Top-up balance in ₹ INR- add credits via instant UPI or cards with 1 credit = 1 paisa (₹0.01).
[3]
Point your agent client- configure baseURL: "https://api.interference.dev/v1" in any standard OpenAI SDK.

Engineered for Autonomous AI Agents

[*]

With < 15ms edge routing overhead, 100% ACID reservation guarantees, and 70+ high-throughput models, Interference powers reliable production compute.

Fig 1.< 15ms P99 Routing Overhead
Fig 2.100% ACID Balance Holds
Fig 3.70+ High-Throughput Models

Zero SDK Modifications

Works with your existing code in TypeScript, Python, Go, or cURL.

agent.ts — @ai-sdk/openai-compatible
TypeScript
import { createOpenAICompatible } from '@ai-sdk/openai-compatible';
import { streamText, tool } from 'ai';
import { z } from 'zod';

const interference = createOpenAICompatible({
  name: 'interference',
  baseURL: 'https://api.interference.dev/v1',
  apiKey: process.env.INTERFERENCE_API_KEY, // sk_live_...
});

const result = await streamText({
  model: interference('claude-sonnet-5'),
  prompt: 'Analyze high-frequency telemetry and execute tool-calls.',
  tools: {
    getLedgerBalance: tool({
      description: 'Fetch user balance in ₹ INR',
      parameters: z.object({ userId: z.string() }),
      execute: async ({ userId }) => ({ balanceCredits: 25000, inr: 250 }),
    }),
  },
});

for await (const delta of result.textStream) {
  process.stdout.write(delta);
}

Transparent Token Pricing

1 Credit = 1 Paisa (₹0.01) • No hidden markup • Real-time D1 ledger tracking

Model Catalog & Relative Throughput1 Credit = 1 Paisa (₹0.01)
Claude Sonnet 5Flagship
in: 25.5 cr / out: 127.5 cr
1M ctx
GPT-5.4Reasoning
in: 31.9 cr / out: 191.3 cr
1M ctx
Gemini 3.7 FlashFast
in: 19.1 cr / out: 95.6 cr
1M ctx
Grok 4.6
in: 25.5 cr / out: 76.5 cr
500k ctx
Kimi K2.7 CodeCoding Agent
in: 12.1 cr / out: 51 cr
262k ctx
Qwen3 Coder
in: 5.7 cr / out: 23 cr
262k ctx
GLM-5.2
in: 17.85 cr / out: 56.1 cr
1M ctx
DeepSeek V4 FlashUltra Low Cost
in: 1.8 cr / out: 3.6 cr
1M ctx
Streaming Throughput (tok/s)Real-time D1 ledger deduction
View top-up presets →

FAQ

Be the first to know when we release new models & gateway features

Get notified for high-throughput releases, price drops, and agent frameworks.