1. Product Introduction
Step Plan is a subscription-based AI service from the StepFun Open Platform designed for high-frequency AI developers. It allows users to access StepFun flagship models at a highly cost-effective price across mainstream coding tools and agent platforms such as OpenClaw, Claude Code, Trae, and Cursor.Core Advantages
- The Step 3.5 Flash model family, proven to be better suited for raising lobsters: ultra-fast inference + hard-core performance = the ideal brain for agents
- Fast model experience by default: all plans provide a unified high-speed model experience.
- 2x the usage of comparable competitor tiers: use it with confidence in agent scenarios.
- Available across platforms: no platform restrictions, with one plan covering mainstream toolchains.
- Continuously expanding model capabilities: Step Plan currently supports
step-3.5-flash-2603andstep-3.5-flash, with more StepFun flagship models to be added over time.
Core Scenarios
OpenClaw: Agent Performance Optimization
- High-speed inference and high-frequency calls with second-level real-time responses
- Deep optimization for agent brain logic
- A flagship model ranked No. 1 by OpenClaw usage
Advanced Coding Development
- Long-context support for handling complex code refactoring with ease
- Repository-level code understanding for precise visibility into the entire engineering context
- End-to-end autonomous engineering to support high-intensity development scenarios
2. Plan Benefits Comparison
Plan benefits table
| Benefits / Plan | Flash Mini (Entry Edition) | Flash Plus (Advanced Edition) | Flash Pro (Professional Edition) | Flash Max (Flagship Edition) |
|---|---|---|---|---|
| Standard Subscription Price | $6.99 | $9.99 | $29 | $99 |
| Applicable Scenarios | Getting started | Daily productivity | Heavy AI use | Professional workloads |
| 5-Hour Limit | 100 prompts (~1,500 requests) | 400 prompts (~6,000 requests) | 1,500 prompts (~22,500 requests) | 5,000 prompts (~75,000 requests) |
| Weekly Limit | 400 prompts (~6,000 requests) | 1,600 prompts (~24,000 requests) | 6,000 prompts (~90,000 requests) | 20,000 prompts (~300,000 requests) |
A “prompt” is our standard billing unit, not a single request. Multiple requests are converted into ‘prompt’ usage to account for varying context lengths and tool usage. Typically, 1 prompt equals 15-20 requests.
3. Supported Models
Currently Available
Step 3.5 Flash 2603- Optimized from Step 3.5 Flash for high-frequency agent scenarios
- Improved token efficiency and faster reasoning speed
- New Low Think Mode to significantly reduce token consumption
- Better compatibility with coding workflows and agent frameworks
- Sparse MoE architecture with 196B total parameters / 11B active parameters
- High-speed inference
- Optimized for agent and coding tasks
- Suitable for agent frameworks such as OpenClaw
Coming Later
We will gradually introduce more StepFun flagship models, including:- Existing voice, image, and video models
- Future new foundation models
4. Supported Tools
You can use Step Plan in mainstream agent tools and coding platforms:Available tools
- Agent tools: OpenClaw, Claude Code, and other agent tools
- AI coding assistants: Cline, Goose, and more
- Developer tools and plugins: Kilo Code, Roo Code, and related integrations
Why Is Step Plan Suitable for OpenClaw?
- Fast response: The Step 3.5 Flash series is optimized for low-latency, real-time agent workflows
- Better suited for task execution scenarios: ideal for long-running and tool-calling agent workflows
- Easier cost control: the combination of subscription pricing and clear limits makes high-frequency agent usage easier to manage
5. Frequently Asked Questions (FAQ)
Basic Questions
Q: What is Step Plan? A: Step Plan is a subscription service launched by StepFun. It uses a 5-hour limit and a weekly limit, and calls models in supported tools and workflows through an API Key. Q: Why emphasize “fast performance”? A: Many similar products separate performance into standard and high-speed versions, and the faster version is usually priced higher. Step Plan aims to make fast response the default experience rather than an expensive add-on. Q: What kind of users is it suitable for? A: It is suitable for high-frequency AI developers, developers using OpenClaw and other agent frameworks, and creators and builders who consume large amounts of tokens in real workflows. Q: What is the difference between Step Plan and calling the API directly? A: Step Plan uses a subscription model. Through the 5-hour limit and weekly limit mechanism, it provides more token usage at a lower price. It also provides usage throttling to help users control costs. Q: What is the difference between a request (Request) and a Prompt? A: Our billing unit is Prompt, which is not equal to a single model call. To standardize consumption across different context lengths and tool calls, multiple model calls are converted into Prompt usage. 1 Prompt ≈ 15–20 model calls. For example, with Step Plan Mini, 100 Prompts equals approximately 1,500 model calls — more than 2x the ~600 calls offered by comparable competing products at the same price point.Plans and Pricing
Q: What is the difference between the different plans? A: The main differences are in the 5-hour limit, weekly limit, and concurrency support. All plans provide the same fast model performance. Q: How long will the promotional price last? A: The promotional price is a limited-time offer. Please refer to official announcements for details. Q: What happens if I exceed the limit? A: After exceeding the 5-hour limit, you need to wait for the next 5-hour cycle to reset. After exceeding the weekly limit, you need to wait for the next weekly cycle to reset. Q: Can unused quota accumulate? A: No. Both the 5-hour limit and the weekly limit reset periodically and will not roll over into the next cycle. Q: How are the weekly limit and 5-hour limit calculated? A: Weekly limit ≈ 5-hour limit × 4. For example, the Flash Mini plan has a 5-hour limit of ~100 Prompts, corresponding to a weekly limit of ~400 Prompts, and so on for other tiers.Model-Related Questions
Q: Which models are currently included? A: Step Plan currently supports Step 3.5 Flash and Step 3.5 Flash 2603, but this is only the beginning. More models are on the way, beyond the Flash series. Q: Will more models be added in the future? A: Yes. Step Plan will gradually expand to include more existing and future StepFun flagship models. Q: Does it support multimodal input? A: The currently available models are Step 3.5 Flash and Step 3.5 Flash 2603. Multimodal capabilities will expand based on each model’s capabilities and future product updates.Feature-Related Questions
Q: Does Step Plan support MCP? A: MCP is not supported at launch. Relevant support is already being planned and is expected to be introduced gradually in future versions.How to Use
Q: How do I use Step Plan? A: After subscribing, obtain an API Key and configure it in a supported tool to use it. Q: Which tools are supported? A: It supports agent tools such as OpenClaw and Claude Code, AI coding assistants such as Trae, Cursor, and Cline, and VS Code plugins such as Kilo Code and Roo Code. Q: Which Base URL does Step Plan use? A: Step Plan has a dedicated endpoint. Please usehttps://api.stepfun.ai/step_plan/v1 instead of the standard API’s https://api.stepfun.ai/v1. Using the wrong URL will cause errors — make sure to use the correct one.
Q: Where can I find the Step Plan integration guide?
A: Please refer to the official Step Plan Integration Guide, which covers configuration for mainstream IDEs and development tools.
Q: What should I do if Cursor / Claude Code or other third-party tools throw errors?
A: Please check the following in order: 1) Confirm that the Base URL is set to the dedicated endpoint https://api.stepfun.ai/step_plan/v1 (not the standard API URL — this is a common mistake); 2) Reconfigure by following the official integration guide; 3) If the issue persists, please provide a specific error screenshot for further assistance.