Farebox

The LLM gateway built for AI agents. One URL. Every model. Pay per request. $FBOX
CA : Gt4XrcDxxn8Ltt9qcvxH7r7Ss1nsHLSBG2t3Cowxpump

[Farebox @Farebox_]

Major free tier update on Farebox

you can now use 18 frontier AI models completely free, no credits required.

Just open the Playground at https://t.co/pLUq0E6TMH, pick any ✦ FREE model, and start chatting. That's it.

What's free right now:

⚡ Via @GroqInc ultrafast responses: Llama 3.3 70B · Llama 3.1 8B · GPT OSS 120B · Qwen 3.6 27B · Groq Compound

🆓 Via @OpenRouter free tier: NVIDIA Nemotron Ultra 550B · Super 120B · Omni 30B · Nano 30B · Nano 12B VL · Nano 9B Poolside Laguna S & XS 2.1 OpenAI GPT OSS 20B Google Gemma 4 26B & 31B Cohere North Mini Code inclusionAI Ling 3.0 Tiny

Every response shows real-time token usage and cost even when it's $0.000. You always know exactly what you're consuming.

Try all 18 free models → https://t.co/cRCBvO5ctx

@GroqInc @NVIDIAAI @poolsideai @OpenAI @GeminiApp @AIatMeta @TheInclusionAI

[Farebox @Farebox_]

This is what healthy infrastructure looks like. $FBOX

Across our active models Claude Opus 4.5, Claude Sonnet 4.5, GPT-4o-mini, Llama 3.3 70B, and more real users are making real requests, and every single one is settling on-chain in USDC.

No simulated volume. No test traffic. Just organic usage growing consistently, request by request.

This is what we track instead of price:

→ Are users returning? Yes.
→ Are models responding reliably? Yes.
→ Is spend increasing over time? Yes.
→ Is the infrastructure holding? Yes.

When usage grows steadily and the system stays stable, everything else follows eventually. The fundamentals are healthy. The platform works. The growth is real.

We'll keep building. The data speaks for itself.

[Farebox @Farebox_]

This is bigger than it looks. $FBOX

MCP is quickly becoming the universal plugin layer for AI the same way HTTP standardized the web. Claude Desktop uses it. Cursor uses it. Windsurf uses it. And now any of those tools can point directly at Farebox and instantly access 71 frontier models through a single endpoint, no custom integration needed.

What makes this significant for developers is the "build once" reality. You write your MCP server using the official Python SDK, connect it to Farebox, and every model GPT-4o, Claude, Gemini, Llama, Mistral, and more becomes available through the exact same interface. Switch models without rewriting your integration. Pay in USDC on Solana. No subscriptions, no per-provider accounts.

For AI agents and autonomous workflows, this is especially useful. MCP gives agents a structured way to call tools and access context. Farebox gives those same agents access to the full frontier model landscape with on-chain billing and transparent usage tracking.

The ecosystem is converging. MCP as the protocol, Farebox as the model layer this combination is worth paying attention to.

[Farebox @Farebox_]

Farebox Supports MCP The Official Python SDK Works With It Natively

The Model Context Protocol is the emerging standard for connecting AI models to context sources, tools, and services. The official Python SDK makes it straightforward to build MCP servers and clients.

Farebox already exposes an MCP endpoint meaning any MCP-compatible client can plug into Farebox and call 71 frontier models through the same protocol that Claude Desktop, Cursor, and other tools use.

Build once on MCP. Get access to every model on Farebox.

→ https://t.co/JZnUL2Up75

[Farebox @Farebox_]

Farebox Supports MCP The Official Python SDK Works With It Natively

The Model Context Protocol is the emerging standard for connecting AI models to context sources, tools, and services. The official Python SDK makes it straightforward to build MCP servers and clients.

Farebox already exposes an MCP endpoint meaning any MCP-compatible client can plug into Farebox and call 71 frontier models through the same protocol that Claude Desktop, Cursor, and other tools use.

Build once on MCP. Get access to every model on Farebox.

→ https://t.co/JZnUL2Up75

[Farebox @Farebox_]

LangGraph Agents Can Route Through Farebox 71 Models, One Endpoint $FBOX

LangGraph is the most widely used framework for building stateful, resilient AI agent pipelines. Cycles, conditionals, human-in-the-loop, persistent state it handles all of it.

It's fully OpenAI-compatible. That means any LangGraph agent can use Farebox as its LLM backend swap the endpoint, keep every graph, node, and edge exactly as is.

Use Claude for reasoning nodes. DeepSeek for code nodes. Gemini for long-context retrieval. All from one Farebox API key, billed per token in USDC.

→ https://t.co/8qvQ3ss2kQ

[Farebox @Farebox_]

Farebox Runs on the x402 Protocol Here's the Repo That Started It $FBOX

x402 is an open HTTP payment standard that lets any server charge for its responses no accounts, no billing cycles, just a 402 status code with payment instructions baked in.

Farebox is one of the first production implementations of x402 on Solana. Every AI inference call you make through Farebox is settled on-chain using this exact protocol.

If you're building anything on x402 payment middleware, agent wallets, HTTP monetization Farebox is the live reference you can call right now.

→ https://t.co/yys05XSX5A

[Farebox @Farebox_]

Build Autonomous AI Agents with Composio × Farebox

Your agent can now think, act, and pay all without human intervention.

Composio gives your agent hands: 250+ integrations to GitHub, Slack, Gmail, Notion, Jira, Linear, and more one unified tool API, zero auth headaches.

Farebox gives your agent a brain: 71 frontier models from Anthropic, OpenAI, Google, xAI, Meta & more one OpenAI-compatible endpoint, billed per token in USDC on Solana.

Plug them together in two lines:
→ Set base_url to api. farebox. fun/v1 in your Composio agent
→ Your agent now reasons through Farebox and acts through Composio

Add x402 and your agent pays for its own inference autonomously no accounts, no invoices, no human in the loop.

The full autonomous agent stack:
✦ Composio tool execution layer
✦ Farebox inference + payment layer
✦ Solana settlement layer

Get started: → https://t.co/dVobfqRSAR

[Farebox @Farebox_]

We’ll be joining the @OOBEonSol on Space and can’t wait to tune in, contribute to the conversation, and connect with others building the agentic commerce ecosystem.

See you there.

[OOBE PROTOCOL @OOBEonSol]

We’re joining Agentic Talks Episode 1 by @xona_agent.

Topic: The State of Agentic Commerce

Co-hosts: @x402 & @SebastianWols17

⏰ Friday, August 14 - 2 pm UTC

Space link⬇️

[Farebox @Farebox_]

@deepseek_ai builds some of the most cost-efficient frontier models in the world. DeepSeek V3 and R1 punch well above their price point strong at code, math, and structured reasoning.

On Farebox, DeepSeek models are among the cheapest per-token options for agents that need reliable technical output without paying Anthropic or OpenAI prices. $FBOX

Great default for coding agents, data extraction, and anything where accuracy matters more than personality. #Farebox

[Farebox @Farebox_]

Eliza Agents Can Now Use Any Model on Farebox 2 Lines of Config

@elizaOS is one of the most powerful autonomous agent frameworks in crypto 18k stars, built for real on-chain agents.

Eliza v1 supports custom OpenAI-compatible endpoints via OPENAI_API_URL. Point it at Farebox and your agent instantly gets:

→ 71 frontier models Claude, Gemini, Grok, DeepSeek, Llama & more
→ No provider lock-in switch models per agent, per task
→ Per-token billing in USDC on Solana → Autonomous payment via x402 agent pays its own inference

Two env vars. Zero code changes: OPENAI_API_URL = https://t.co/rBkzJeCMKt OPENAI_API_KEY = sk-fbx-... from dashboard

Your @elizaOS agent now thinks with any frontier model and pays for it on-chain no subscriptions, no invoices, no human in the loop.

[Farebox @Farebox_]

That's what Solana Name Service (SNS) integration means for Farebox.

Here's how it works 👇

Right now:

Connect your wallet
Farebox sees your raw address: 7xKXt...9mNp
That's what shows everywhere billing, dashboard, payments
With SNS:

Connect your wallet
Farebox checks on-chain: "does this wallet have a .sol name?"
It does → alice.sol
That's what shows everywhere instead
Your identity. Human-readable. On-chain verified.

It also works for payments.

Want to top up another user?

Instead of asking them to copy-paste their 44-character address just type bob.sol

Farebox resolves it → sends USDC to the right wallet automatically.

Why does this matter?
Most AI APIs have zero on-chain identity.

Farebox users will have verified .sol names tied to their usage, billing, and relay node activity.
$FBOX

[Farebox @Farebox_]

Farebox × @sns Solana Name Service Coming Soon

We're integrating SNS into Farebox, bringing human-readable .sol identities to every part of the platform.

Here's what's coming:

→ Dashboard Identity your wallet shows as you.sol, not a 44-char address
→ Relay Node Names node operators identified publicly by their .sol name verifiable, trustless
→ Public Profiles browse any operator at https://t.co/Xf2X0ndRjB, see their stats, uptime & earnings
→ Payment Resolution send USDC to .sol names directly, no address copy-paste needed

The entire x402 payment flow stays on-chain. SNS just makes every actor in it readable by humans.

Farebox is already Solana-native. This is the identity layer it was missing.

[Farebox @Farebox_]

@AIatMeta Llama 4 Maverick and Scout are Meta's open-weight frontier models and among the most capable open models available today.

Maverick for complex reasoning and agent tasks. Scout when you want a leaner, faster variant without sacrificing too much capability.

Both are available on Farebox. Open-weight models, closed-source convenience routed, billed, and streamed through one API. $FBOX

[Farebox @Farebox_]

@grok is built by xAI with real-time awareness and strong reasoning baked in. It's opinionated, fast, and doesn't hedge when it doesn't need to.

Grok 4 is available on Farebox well-suited for agents that need direct answers, financial reasoning, or tasks where verbose safety wrapping slows things down.

One endpoint. No account with xAI needed. Pay per token in USDC on Solana. $FBOX

[Farebox @Farebox_]

@GeminiApp Gemini 2.5 Pro has the largest context window of any model on Farebox up to 1M tokens. That means entire codebases, full legal documents, hours of transcripts all in one call.

Gemini 2.5 Flash is its speed-optimized sibling: near-instant responses with strong reasoning at a much lower token cost.

If your agent needs to process massive context without chunking, Gemini is the move. Available on Farebox, billed per token in USDC. $FBOX

[Farebox @Farebox_]

📢 Farebox Development Update

We know the token price doesn't reflect what we're building. But we haven't stopped not for a day. Here's the proof.

What we just shipped:
→ 36 built-in AI Skills live summarize, translate, code review, on-chain wallet analysis, trader tools & more. Callable via REST or MCP right now.
→ 71 frontier models across 13 providers (Anthropic, OpenAI, Google, xAI, Meta, DeepSeek & more) one endpoint, one API key.
→ Full security audit completed. HIGH severity issues patched. Your funds and data are safe.
→ 18+ free models available no deposit needed to start.

What we're building next:
→ Solana Name Service (SNS) integration your wallet becomes you.sol across the platform
→ Skill usage analytics see which skills developers use most
→ Anti-abuse rate limiting stronger, more resilient infrastructure
→ Deep integrations with major agent ecosystems: Eliza, LangGraph, CrewAI, Composio & more

To every holder:

A low token price is not a failed project. It's a market that hasn't caught up yet.

We're not building hype. We're building infrastructure an AI gateway that's truly on-chain, pay-per-token, no accounts, no subscriptions, no monthly seats. Every feature we ship is live and usable at https://t.co/pLUq0E6TMH today.

Real utility. Real development. We're not going anywhere.

[Farebox @Farebox_]

The Playground That Lets You Test 65 Models Side by Side Before You Pay a Cent

One underused feature of Farebox is the free playground 4 models available with zero wallet setup, letting developers compare outputs across providers before committing to a stack. Most teams pick a model once during prototyping and never revisit the decision. The cost of switching later prompt reformatting, evals, re-integration is high enough that the initial choice sticks even when better options exist.

Having GPT-5, Claude Sonnet 4.5, Gemini 2.5 Pro, Grok 4.5, and DeepSeek V3 all reachable from one endpoint changes the calculus. You can run the same prompt across all of them in a single session, compare quality and cost per token, and make an informed architecture decision before your first dollar leaves the wallet. That's a workflow most AI teams don't have today. $FBOX #Farebox

[Farebox @Farebox_]

@AxiomBot Yes and if that receipt is signed by the browser process and anchored on-chain, you get a full trustless audit trail: Camofox proves the browse, Farebox proves the inference, Solana proves the payment. No step requires trust.

[Farebox @Farebox_]

Build AI Agents That Can Actually Browse the Web Without Getting Blocked

Most AI agents fail the moment they hit Cloudflare, CAPTCHA, or bot detection. Camofox fixes that.

Camofox is an open-source stealth headless browser built specifically for AI agents it bypasses Cloudflare, fingerprint detection, and anti-scraping systems. Drop-in replacement for Puppeteer and Playwright, but designed to be invisible. $FBOX

Pair it with Farebox and you have a complete autonomous research agent:

→ Camofox visits any site without being blocked scrapes data, reads pages, captures content
→ Farebox processes what it finds summarize, extract structured data, analyze, generate reports
→ x402 lets the agent pay for every inference call autonomously in USDC no human needed

Real example: Agent monitors competitor pricing
→ Camofox opens 50 pages undetected
→ Farebox extracts the data and writes a structured report
→ all settled per-token on Solana

Two tools. One autonomous pipeline. No API keys expiring. No bot blocks. No monthly seats.

→ https://t.co/zhykadAo7A

[Farebox @Farebox_]

The models that started it all now accessible through Farebox with no monthly seat, no rate-limit account, no setup friction.

@OpenAI GPT-4o for multimodal tasks. o3 for deep, multi-step reasoning. GPT-4o-mini when you want OpenAI quality at a fraction of the cost.

Swap your base_url to Farebox and your existing OpenAI code works instantly. Same SDK. Same format. Pay per token in USDC only what you actually use. $FBOX

[Farebox @Farebox_]

@claudeai is the model you reach for when accuracy and safety actually matter.

Claude Sonnet 4.5, Opus 4.5, and Haiku 4 are all available on Farebox each tuned for a different tradeoff between speed, depth, and cost.

Use Opus for deep reasoning and long documents. Sonnet for production agents that need reliable, grounded answers. Haiku when you need Claude-quality output at near-instant speed.

All routed through one OpenAI-compatible endpoint, billed per token in USDC. $FBOX