Paperclip + Anthropic
Host Claude agents on Paperclip — Fable 5.1, Opus 5, Sonnet 5, and Haiku 4.5 with 1M-token context, on a managed runtime from $3.99/mo.
Claude agent hosting, in one answer
Anthropic sells tokens and, since April 2026, Managed Agents on its own infrastructure. Paperclip gives your Claude agents the rest of the production stack: a supervised runtime that restarts crashed processes, persistent memory between runs, per-run monitoring, and model routing across providers. Bring your existing Anthropic API key and deploy in five minutes — instruction-heavy agents run for hours without drifting, and one hosted bill replaces the glue code.
Claude agent models and prices (September 2026)
Newest flagship for demanding reasoning and long-horizon agentic work — the step up when Opus 5 at higher effort still falls short. $10.00 in / $50.00 out per 1M tokens, 1M-token context, and a 2.5% cache-read rate ($0.25/MTok).
Flagship long-horizon reasoning. $5.00 in / $25.00 out per 1M tokens with a permanent 1M-token context.
Balanced default at $2.00 in / $10.00 out per 1M tokens, now a permanent price point. Recommended for most agents.
High-volume tier at $1.00 in / $5.00 out per 1M tokens with 200K context. For routing and extraction at scale.
How to deploy Claude agents
Get your Anthropic API key
Go to console.anthropic.com → API Keys → Create key.
Add it to HostAgentes
Dashboard → Agent Settings → Environment Variables → Add ANTHROPIC_API_KEY.
Select your model
Choose Fable 5.1, Opus 5, Sonnet 5, or Haiku 4.5 from the model dropdown — or let model routing pick per request.
Deploy
Click deploy. Your Claude-powered agent is live with persistent memory, auto-restarts, and run-level monitoring.
Claude agent capabilities
Tool use
Native function calling with reliable multi-step argument handling.
Extended thinking
Available on every Claude model for deliberate multi-step reasoning.
1M-token context
Fable 5.1, Opus 5, and Sonnet 5 take 1M tokens; Haiku 4.5 stays at 200K.
Prompt caching
Cache reads bill at 10% of the input rate on most Claude models — and at just 2.5% ($0.25/MTok) on Fable 5.1.
Cost optimization tip
Route high-volume tasks to Claude Haiku 4.5 at $1.00 per 1M input tokens, keep long-horizon reasoning on Sonnet 5 or Opus 5, and reserve Fable 5.1 for the runs where Opus 5 at higher effort still falls short — its 2.5% cache-read rate ($0.25/MTok) softens the $10/$50 sticker price on long-running agents. Cache reads bill at 10% on most Claude models, and batch requests take another 50% off. HostAgentes lets you switch models per agent — no redeployment needed.
HostAgentes vs Claude Managed Agents
Managed Agents keeps the execution harness inside Anthropic's stack and ties you to Claude models. Paperclip is provider-neutral: the same agent can route runs to Claude, GPT-5.6, Gemini, or Mistral, with memory, restarts, and monitoring handled the same way regardless of model.
Model lineup and prices verified against Anthropic's public pricing documentation and the Claude Managed Agents documentation in September 2026. Provider prices change; re-check before committing volume.
Frequently asked questions
Doesn't Anthropic already offer Claude Managed Agents?
Anthropic's Managed Agents (launched April 2026) run pre-built agent harnesses on Anthropic-managed infrastructure for long-running and asynchronous tasks. That execution layer still leaves memory strategy, restarts, monitoring, and multi-provider routing to your team. HostAgentes is the managed runtime around any Claude model: bring your Anthropic API key and Paperclip handles process supervision, persistent memory, auto-restarts, per-run monitoring — and routing to OpenAI, Gemini, or Mistral when a task suits another model.
What is Claude Fable 5.1 and should my agent use it?
Fable 5.1 (model ID claude-fable-5-1) is Anthropic's newest flagship, positioned for demanding reasoning and long-horizon agentic work — the documented step up when your evals on Claude Opus 5 at higher effort still fall short. It costs $10 per 1M input tokens and $50 per 1M output, with a 1M-token context window, 128K max output, and a uniquely low 2.5% cache-read rate ($0.25/MTok) that rewards long-running agents. Start on Sonnet 5 and move to Fable 5.1 when evals justify it.
Which Claude model should my agent use?
Start with Claude Sonnet 5 at $2.00 per 1M input tokens — it handles most production agent workloads and its price is now permanent. Route high-volume simple tasks to Haiku 4.5 ($1.00 input, 200K context), long-horizon planning to Opus 5, and the hardest long-running evaluations to Fable 5.1. HostAgentes lets you switch models per agent without redeploying.
How much does it cost to run Claude agents?
A typical agent run spends about 1,000 input and 100 output tokens, so one million runs cost about $1.50 on Haiku 4.5, $3.00 on Sonnet 5, and $7.50 on Opus 5 at September 2026 list prices — before the 50% batch discount and 10% cache-read rate.
Can I switch Claude models without redeploying?
Yes. Change the model per agent from the HostAgentes dashboard at any time, or use model routing to send high-volume runs to Haiku and long-horizon planning to Opus. No redeployment needed.
Deploy a Claude-powered agent in 5 minutes
Bring your Anthropic API key. We handle the rest.
Get Started