Build Cost-Efficient AI Agents: Use MiniMax M2.5 in OpenClaw (Clawdbolt) via Novita AI

AI agents are rapidly transforming how we interact with applications — from chat interfaces to fully automated workflows. Among open-source agent frameworks, OpenClaw has emerged as a flexible, multi-model scaffold that connects messaging platforms with powerful LLMs. Pairing it with advanced models like MiniMax M2.5 unlocks efficient, cost-effective, and high-performance AI assistants capable of complex reasoning, code generation, automation, and more — all within a low-cost footprint.

In this post, we explain what OpenClaw is, why MiniMax M2.5 is an exciting choice, and how to actually integrate MiniMax M2.5 into an OpenClaw deployment. This is particularly valuable for Novita AI users who want to build capable AI agents that scale affordably.

What Is OpenClaw?

OpenClaw (previously known as Clawdbot or Moltbot) is an open-source AI assistant framework designed to turn language models into versatile agents. It:

• Integrates various AI models, routing requests to the appropriate model backend.
• Hooks into popular messaging platforms such as WhatsApp, Telegram, Discord, iMessage — enabling direct conversational control of AI agents.
• Enables task automation and integrated skills like file read/write, web search, and workflow tooling through extensible plugins and “skills”.

OpenClaw’s agent model isn’t just about dialog — it’s about executing actions, reasoning over tasks, and orchestrating workflows across platforms.

What is MiniMax M2.5 and Why use MiniMax M2.5?

MiniMax M2.5 is a next-generation AI model that shines in real-world productivity tasks:

• It showcases a balance of strong output quality and efficient performance, making it suitable for agent workloads like coding, multi-step reasoning, and instruction following.
• Compared with larger commercial models, M2.5 delivers high throughput and low cost, which matters when integrating into production agent systems like OpenClaw.
• Its architecture is designed for agent-oriented training, meaning it handles action sequences, long dialogues, and tool usage more gracefully than many generative backends.

MiniMax vs. Kimi K2.5 vs. GLM-5: Which One Should You Choose?

When selecting a default model for OpenClaw, developers commonly compare MiniMax M2.5, Kimi K2.5, and GLM-5. All three operate in a similar capability tier and are viable for agent-based workflows, but they emphasize slightly different strengths.

• MiniMax M2.5 tends to be the most balanced choice for structured agent execution. If your OpenClaw deployment focuses on workflow automation, document generation, coding tasks, or tool-heavy interactions, MiniMax M2.5 offers a pragmatic blend of reasoning ability, speed, and cost efficiency.
• GLM-5 is often preferred when long-context reasoning is the priority. If your agents frequently process large documents, logs, or extended knowledge bases in a single session, GLM-5’s extended context handling can be advantageous.
• Kimi K2.5 is particularly compelling in multimodal or visually grounded scenarios. If your workflow involves interpreting screenshots, UI elements, charts, or image-based inputs in addition to text reasoning, Kimi K2.5 may provide a stronger experience.

In many real-world OpenClaw deployments, the question is not “Which model is the absolute best?” but rather “Which model best fits the dominant workload?” For developers building scalable, cost-sensitive production agents on Novita AI, MiniMax M2.5 often emerges as the most balanced and deployment-friendly default — while still leaving room to route specialized tasks to other models if needed.

API Pricing on Novita AI (Per 1M Tokens)

Model	Input Cost	Cache Read Cost	Output Cost
MiniMax M2.5	$0.30	$0.03	$1.20
Kimi K2.5	$0.60	$0.10	$3.00
GLM-5	$1.00	$0.20	~$2.55

How to Use MiniMax M2.5 in OpenClaw (6 Steps)

Preparation

Ensure you have:

• A machine that can access the internet (Windows users often use WSL2)
• Node.js 22+
• Deployment choice: OpenClaw can run on a local machine or on a cloud VM/container, depending on how you want to expose the gateway and manage ops.

Step 1: Get your API Key

In Novita, create an API key (it’s typically shown once—store it immediately). The reference flow is:

1. Visit https://novita.ai, Create/login to your Novita account
2. Go to API Keys
3. Click Add new key
4. Copy and save it right away

Step 2: Install OpenClaw

Run the one-line installer for your platform:

• macOS / Linux / WSL2

curl -fsSL https://openclaw.ai/install.sh | bash

• Windows (PowerShell)

iwr -useb https://openclaw.ai/install.ps1 | iex

Step 3: Run the OpenClaw configuration

Start the OpenClaw configuration wizard:

openclaw config

Next, we will configure the Workspace, Model, Gateway, and Channels (using Telegram as an example) step by step:

1. Workspace

• Where will the Gateway run? → Choose Local.
• Select sections to configure → Workspace

2. Model

• Select sections to configure → Model
• Model/auth provider → Skip for now
• Filter models by provider → All providers
• Default model → Enter model manually
• Default model → novita/minimax/minimax-m2.5

novita/minimax/minimax-m2.5 follows the standard novita/<model-id> naming convention, where novita is the provider prefix and the remaining portion represents the exact model ID as displayed on the Novita platform. If you’d like to use a different model available on Novita, simply replace minimax/minimax-m2.5 with the corresponding model ID shown in the Novita console—while keeping the novita/ prefix unchanged.

3. Gateway

Keep defaults for a first run, just press Enter to continue

4. Channels

• Click Configure/Link
• Select the channel Telegram (Bot API) and enter your Telegram Bot Token (see the guide below to obtain it from @BotFather)
• Click Finish
• Set up the DM access policy and select Pairing

❓How to get Telegram Bot Token: Find @BotFather on Telegram and send the command /newbot to create a new bot. After completing the setup process, you will receive a bot token.

Step 4: Use MiniMax M2.5 via Novita AI

So far, the API has not been configured yet. Next, you can follow the steps below to complete the setup and start using Novita’s cost-effective API.

Open your OpenClaw config file:

open ~/.openclaw/openclaw.json

Copy the configuration snippet below into your config file. You only need to modify one field: <Your API Key> → replace it with your Novita API key.

"models": {
  "providers": {
    "novita": {
      "baseUrl": "https://api.novita.ai/openai/v1",
      "apiKey": "<Your API Key>",
      "api": "openai-completions",
      "models": [
        {
          "id": "minimax/minimax-m2.5",
          "name": "Minimax M2.5",
          "reasoning": false,
          "input": [
            "text"
          ],
          "cost": {
            "input": 0,
            "output": 0,
            "cacheRead": 0,
            "cacheWrite": 0
          },
          "contextWindow": 204800,
          "maxTokens": 131100
        }
      ]
    }
  }
},

Step 5: Start the Gateway

Install and restart the OpenClaw gateway:

openclaw gateway install
openclaw gateway restart

Step 6: Pair Telegram

At this point, your Telegram account is not yet paired with the bot you created. You need to open the bot you created and send /start. The bot will give you pairing code, then approve it in a new terminal:

openclaw pairing approve --channel telegram <your-pairing-code>

Conclusion

Integrating MiniMax M2.5 into OpenClaw unlocks a powerful blend of conversational AI, task automation, and agent execution — all at accessible cost. Whether you’re building internal tools, customer service assistants, or productivity bots, this pairing offers both performance and scalability.

Novita AI users can leverage this setup to build differentiated applications that deliver real value in real-world workflows — faster, cheaper, and with high quality results.

Novita AI is an AI cloud platform that offers developers an easy way to deploy AI models using our simple API, while also providing the affordable and reliable GPU cloud for building and scaling.

Frequently Asked Questions