Built for WEB4
The Agent Economy Infrastructure
& All Agents 
One API. Any Official Model. No signup, no humans in the loop. Agents connect their wallet, fund themselves, and work 24/7.
You set the mission. Your agents do the rest — register, pay, reason, earn, repay. Forever.
No email. No password. No KYC. Wallet address = account.
Models & Pricing
One API For Any Model · Official API Direct · USDT / USDC on Ethereum
Claude Fable 5.1
Anthropic's most capable model — the new CEO. Next-generation reasoning for the hardest problems and multi-hour agentic missions, always-on adaptive thinking, 1M context. Same $10/$50 as Fable 5 with cache reads at $0.25/M (a quarter of Fable 5's rate). Forced tool_choice is not supported upstream — use auto plus an explicit instruction.
Claude Opus 5
The new enterprise flagship for complex agentic coding and enterprise work. Near-Fable capability at half the price — with a May 2026 knowledge cutoff, the freshest knowledge in the Claude lineup.
GPT-5.6 Sol
OpenAI's new frontier flagship. Successor to GPT-5.5 with a new max reasoning tier and Ultra subagent mode — state-of-the-art for the hardest long-horizon agentic coding and tool orchestration.
GPT-5.6 Terra
GPT-5.5-class quality at 40% of the price. The balanced everyday frontier model — strong reasoning, coding, and tool use for agents that don't need flagship cost.
Gemini 3.1 Pro
Google's flagship multimodal model. Best-in-class reasoning, 1M context, exceptional at scientific research and data analysis.
Claude Sonnet 5
The best combination of speed and intelligence. Near-Opus quality on coding and agentic work at Sonnet cost — $2/$10 standard pricing.
GPT-5.6 Luna
The budget agentic workhorse. Frontier-family reasoning at $0.20/M input with 90%-off cache reads and 30-minute prompt caching — built for high-volume agent fleets.
Gemini 3.8 Flash
Google's agentic Flash — native function calling (streaming + non-streaming), coding and multi-step agent pipelines at Flash cost, 1M context. Intro price $0.75/$3.75 through 2026-12-31; $1.50/$7.50 from 2027-01-01.
GPT-OSS-120B
Production-grade reasoning powerhouse. Near o4-mini performance with full chain-of-thought, function calling, and configurable reasoning depth.
GPT-OSS-20B
Lightning-fast inference for latency-sensitive apps. Runs within 16GB memory — perfect for edge deployment and specialized tasks.
GPT-5.4
OpenAI frontier — coding, agents, computer use, 1M context. The recommended default for most tasks.
Claude Opus 4.7
Legacy Opus kept available for workflows pinned to 4.7 behaviors. Two generations back — prefer Claude Opus 5 for new work at the same price.
Claude Sonnet 4.6
The battle-tested previous-gen Sonnet. Kept available for workflows tuned to 4.6 behaviors — Claude Sonnet 5 offers more intelligence at a lower price.
GPT-5.5
The battle-tested agentic model. Proven at long-horizon coding, computer use, and tool orchestration — kept at full strength for workflows tuned to 5.5; GPT-5.6 Terra offers similar quality at 40% of the price.
GPT-5.4 Mini
OpenAI's proven mini model — coding, computer use, and subagents. A dependable known quantity for high-volume workloads; GPT-5.6 Luna is the newer budget-agent pick.
Claude Opus 4.8
Bench strength: the proven Opus flagship kept warm for second opinions and workflows tuned to 4.8 behaviors — same $5/$25 as Claude Opus 5.
Gemini 3.1 Flash-Lite
Google's fastest model — 381 tokens/sec, 1M context. The pure-throughput pick for real-time agents and high-volume pipelines; everyday low-cost work now runs on Luna.
Claude Fable 5
Bench strength: the proven Fable flagship kept warm for workflows tuned to Fable 5 behaviors — and for forced tool_choice, which Fable 5.1 no longer accepts. Same $10/$50; cache reads $1/M.
Gemini 3.7 Flash
Bench strength: the proven agentic Flash kept warm for pipelines tuned to 3.7 behaviors. Intro price $0.75/$3.75 through 2026-12-31; $1.50/$7.50 from 2027-01-01.
Why wallet = account?
Designed for the Agent Economy. Zero human gatekeeping.
Quick Start
Three doors. One key. All models, official APIs.
from openai import OpenAI
client = OpenAI(
base_url="https://clawapi.org/api/v1",
api_key="sk-claw-YOUR_KEY"
)
response = client.chat.completions.create(
model="claude-opus-4-8",
messages=[{"role": "user", "content": "Hello from ClawAPI!"}]
)
print(response.choices[0].message.content)
# GPT-5.6 + tools + reasoning? Use /v1/responses (see below)# Use any client that speaks Anthropic — including Claude Code CLI:
export ANTHROPIC_BASE_URL=https://clawapi.org/anthropic
export ANTHROPIC_API_KEY=sk-claw-YOUR_KEY
claude --model gpt-5.4 # any of the ClawAPI models
# Or the Anthropic SDK directly:
from anthropic import Anthropic
client = Anthropic(
base_url="https://clawapi.org/anthropic",
api_key="sk-claw-YOUR_KEY"
)
msg = client.messages.create(
model="claude-sonnet-4-6", # ClawAPI short ID — stable across versions
max_tokens=200,
messages=[{"role": "user", "content": "Hello"}],
)# GPT-5.6 family: reasoning + function tools runs here (upstream requirement)
from openai import OpenAI
client = OpenAI(
base_url="https://clawapi.org/api/v1",
api_key="sk-claw-YOUR_KEY"
)
r = client.responses.create(
model="gpt-5.6-luna",
input="What's the ETH balance of vitalik.eth?",
tools=[{"type": "function", "name": "get_eth_balance",
"parameters": {"type": "object",
"properties": {"address": {"type": "string"}}}}],
)
print(r.output_text)sk-claw-* key works on every endpoint. Pick by client convenience — every model works on both protocol doors, with one exception: GPT-5.6 family with reasoning + tools must use /v1/responses. cache_control + extended thinking are preserved on the Anthropic endpoint. Forced tool_choice (required / a named function) is not available on claude-fable-5-1 — Anthropic removed it on Fable 5.1, so you get a clear 400, never a silent downgrade to auto. Use auto + an explicit instruction, or claude-fable-5 / claude-opus-5.Use with ClawCompany
Three steps. Under 60 seconds. Your model workforce is ready.
Fable 5.1 is your CEO. Sonnet 5 is your CTO. GPT-5.6 Luna is your CFO. One key runs them all.