Compare · camelRun by camelAI

Claude Managed Agents alternative

One agent, research-agent, runs on Claude Sonnet 5.5, then Claude Opus 5.5, then GPT-6 Luna, then Gemini 3.8 Flash, then DeepSeek V4 Pro, then Kimi K3, then Your own model server, on run.camelai.com and on your own servers at localhost:8790, keeping its history in the same conversation each time it changes model.

Run your AI agent on Claude, GPT, Gemini or your own model, on our cloud or yours.

camelRun by camelAI runs your agent's loop and keeps its history and files. Change its model between runs and the conversation carries over.

Setting up an AI agent on Claude Managed Agents and on camelRun

Claude Managed Agents has you create an agent, an environment and a session. On camelRun you name the agent with your own id, its key, so one file makes it the first time, finds it after that, and runs it.

Claude Managed Agents setup
Create an agent

Claude Managed Agents: Model, system prompt, tools, MCP servers and skills, saved under an ID.

research-agent/index.ts
import { Agents } from "@camelai/run";

const agents = new Agents();

// Built-in tools the runtime answers itself: web search and page reading.
const agent = await agents.upsert("research-agent", {
  model: "openrouter/openai/gpt-6-luna",
  instructions: "You are a careful researcher. Search the web, read the best sources with web_fetch, " +
    "then answer in a few short paragraphs. Cite every claim with a numbered [n] and list the URLs under Sources.",
  builtins: ["web_search", "web_fetch"],
});

const question = process.argv.slice(2).join(" ") || "What changed in the most recent Node.js LTS release?";
for await (const part of agent.stream(question)) {
  if (part.type === "text") process.stdout.write(part.text);
  if (part.type === "tool_call") console.log(`→ ${part.name}(${JSON.stringify(part.arguments)})`);
}
console.log();
await agents.close();

Step 01The model, instructions and tools go into one upsert. Put anthropic/claude-sonnet-5-5 on line 7 to stay on Claude.

The research agent on GitHub
run/examples/quickstarts/research-agent

npm install

export CAMELAI_API_KEY=art_...

npm start -- "What changed in the most recent Node.js LTS release?"

> start

> tsx index.ts What changed in the most recent Node.js LTS release?

→ web_search({"query":"site:nodejs.org/en/blog/release most recent Node.js LTS release"})

→ web_fetch({"url":"https://nodejs.org/en/blog/release/v24.21.0"})

The latest Node.js LTS release is 24.21.0 ("Krypton"), released September 8. [1]

…

Sources

[1] https://nodejs.org/en/blog/release/v24.21.0

run/examples/quickstarts/research-agent

npm install

export CAMELAI_API_KEY=art_...

npm start -- "What changed in the most recent Node.js LTS release?"

> start

> tsx index.ts What changed in the most recent Node.js LTS release?

→ web_search({"query":"site:nodejs.org/en/blog/release most recent Node.js LTS release"})

→ web_fetch({"url":"https://nodejs.org/en/blog/release/v24.21.0"})

The latest Node.js LTS release is 24.21.0 ("Krypton"), released September 8. [1]

…

Sources

[1] https://nodejs.org/en/blog/release/v24.21.0

Your control plane

Host your whole AI agent runtime

camelRun's runtime is open source. One container and Postgres run the agent loop and keep its history and files on servers you own, with the same API as run.camelai.com.

PartClaude Managed Agents, self-hosted sandboxcamelRun, self-hosted
Agent loopAnthropicYour servers
Tool callsYour machinesYour servers, or your own code

Checked October 22, 2026.Anthropic's docs

Point CAMELAI_BASE_URL at it and the same code runs unchanged, on your own model and search keys.

docker compose up -d

export CAMELAI_BASE_URL=http://localhost:8790 # http://runtime:8790 from the Compose network

Channels, built in

Put your AI agent in Slack, Discord, Telegram and GitHub

Ask the research agent from a Slack thread, a Discord DM, a Telegram chat or a GitHub issue. Connect a channel in the console or with one API call, and camelRun takes the messages in and sends the replies.

Channels in the docs
One conversation, one research agent
  • SlackA thread started by an @mention, or a DM
  • DiscordA DM, or a channel or thread where your bot is @mentioned
  • TelegramA chat
  • GitHubA pull request or an issue
  • WebhooksWhatever your request names, like a ticket

Claude Managed Agents and camelRun prices, side by side

Both bill agent time only while an agent is working.

Runtime fee, one hour of work
Claude Managed AgentsIncludes the session's cloud sandbox. Idle time isn't billed.$0.08per session-hour
camelRunYour tools run in your own code. Idle and sleeping agents cost only storage.$0.01per agent-hour
RateClaude Managed AgentscamelRun
TokensClaude's API pricesThe provider's price, no markupOn your own provider keys, nothing from camelRun: the provider bills you.
Web search$10 per 1,000 searches$0.001 to $0.007 per search

Storage $0.10 per GB-month. Prepaid. No subscription, no seats, no per-agent fee.

Checked October 22, 2026.Anthropic's pricingcamelRun pricing

Claude Managed Agents questions

About Claude Managed Agents

Does Claude have managed agents?

Yes. Claude Managed Agents is Anthropic's hosted agent service, in public beta since April 8, 2026. Anthropic runs the agent loop for Claude 4.5 and later models, with a sandbox for each session. camelRun by camelAI is a hosted runtime for AI agents that runs Claude and models from other providers.

How expensive are Claude managed agents?

Claude Managed Agents bills Claude's token prices plus $0.08 per session-hour (its sandbox included) while a session is running, and $10 per 1,000 web searches (Anthropic's pricing page, October 22, 2026). camelRun charges $0.01 per agent-hour while an agent is in a run and $0.10 per GB-month of storage; tokens on your own provider keys cost $0 extra.

Can you self-host Claude Managed Agents?

In part. Its self-hosted sandboxes run the agent's tool calls on your machines, while Anthropic keeps running the agent. You can run the whole camelRun runtime yourself: one container and Postgres run the agent loop and keep its history and files, with the same API as run.camelai.com.

About camelRun

Is there an open source alternative to Claude Managed Agents?

Yes: camelRun. Its runtime is open source under the AGPL, and its SDKs, CLI and UI kits are MIT, so you can use them in any app. Use the hosted runtime at run.camelai.com, or run the same runtime yourself with Docker.

Can I keep using Claude on camelRun?

Yes. Claude Sonnet 5.5 is the default model on run.camelai.com, on our keys or your own Anthropic, Amazon Bedrock or OpenRouter key. Move any agent to GPT, Gemini, DeepSeek, Kimi or a model server of your own between runs; its history carries over.

How do I move an agent from Claude Managed Agents to camelRun?

Give it a key, the same instructions and a model, and put its tools in your code or behind an MCP server. To bring a conversation along, turn its events into Messages API messages: importMessages takes those as they are, tool calls and thinking included.

Bring in existing conversations

Should I use Claude Managed Agents or camelRun?

If your agent needs its own Linux container with a shell and package installs, and Claude is the only model you need, Claude Managed Agents gives every session one. If you want to pick the model for each agent, run the whole runtime on your own servers, or reach people in Slack, Discord, Telegram and GitHub, camelRun is built for that.

Bring your Claude agent to camelRun

Agent
Your own id, like research-agent
Model
Claude, GPT, Gemini or your own
Runtime
Our cloud or yours
Set up with your coding agent

Read https://run.camelai.com/SKILL.md and set up camelRun in this project. Ask me what the agent should do.

Paste into Claude Code, Codex, Cursor or any coding agent. It asks what your agent should do, then builds it.