How do I publish and manage my own Mule Agents on MuleRun?

As the world’s first AI Agent marketplace, MuleRun not only provides users with “plug-and-play” AI tools, but also builds a full chain of services for developers from Agent development, on-boarding to revenue management through exclusive features such as Creator Studio. Whether you are a technology developer, an industry expert, or a creator with specific experience, you can package your professional abilities into Mule Agents and realize “skills for cash”. The following is the detailed process of publishing and managing Mule Agents in MuleRun, combining platform rules and practical skills to help you efficiently start the journey of AI tool creation.

I. Pre-preparation: Meet the basic conditions for publishing

(A) Complete account registration and activation

1. Register an account: visit the official website of MuleRun ( https://mulerun.com/), and complete the registration by email or third-party account (e.g. Google, GitHub); if you have an invitation code (available through Discord community, blogger sharing, Badger watchers’ community, etc., and the new user can get 1000 initial points for registration), you can fill it out when registering to unlock additional benefits. If you have an invitation code, you can fill it in when registering to unlock additional benefits.
2. Upgrade creator status: Login and enter personal center, click “Become a creator”, submit basic information (e.g. status type: independent developer/team/enterprise, field of expertise: content creation/automation/game assistance, etc.) to complete the creator qualification certification. – Individual users need to provide proof of skills (such as past project cases, technical certificates), team / enterprise users need to add organizational information, the platform will complete the review within 1-3 working days.
3. Familiarization with development tools: through the Creator Studio backend (accessed through the “Creator” portal on the official website navigation bar), you can learn about the development support provided by the platform: including one-click call interfaces for 15 mainstream large models (e.g. GPT-4, Claude 3, Stable Diffusion) The platform also provides a multimodal API (text/image/video generation), workflow templates (e.g., automation script templates, content generation templates). It is recommended to try out the “Beginner’s Development Toolkit” to familiarize yourself with the Agent packaging process.

(ii) Define the positioning of Agent and prepare materials

1. Determine the core value of Agent: Define the target users (e.g., “e-commerce operators”, “3D designers”, “hand game players”) and the core pain points to be solved (e.g., “batch generate product literature”, “batch generate product literature”, “batch generate product literature”). “batch generation of product copy”, “rapid production of hand puppet models”, “automatic completion of daily tasks for games”), and need to explain in the application materials “the difference with the Differences with competing products” – for example, “3D Desktop Character Creation Agent” needs to emphasize that it “supports two-step operation of photo uploading → hand puppet model generation, and does not require professional modeling skills. Instead of simply repeating the functionality of existing tools, you should emphasize “supports two-step operation of photo upload → model generation, no professional modeling skills required”.
2. Prepare demo and documentation:
   • Demo production: After completing the development of Agent, record a demo video (1-3 minutes, showing the whole process of “input parameters→starting task→outputting result”) or generate sample test results (e.g., 3D model files, automation task reports) to ensure that the effect is intuitive and verifiable;
   • Explanation document: contains “functional description (such as the types of tasks supported, input and output formats), use scenarios (such as “suitable for e-commerce sellers to batch optimize the product image”), parameter descriptions (such as “the need to enter the image resolution, style type “), precautions (such as “single generation takes about 2 minutes”)”, the document should be concise and clear, to help reviewers quickly understand the Agent value.
3. Plan for profitability: Choose pricing based on Agent type – either on a “single-use credit” basis (e.g., 50 credits per 3D modeling Agent, about $0.50) or on a “subscription basis” (e.g., Automated Office Agent, about $0.50). “(e.g., 200 points per month for an automated office agent), and the platform will share the points with the creators according to the amount of users’ usage (the share ratio increases with the level of the creators, and a high-quality agent can get up to 80% of the points).

Agent development and launch: from tool packaging to review and launch

(I) Development and Packaging: Lowering the Technical Threshold with Platform Tools

1. Select the development method:
   • Low-code template development (suitable for non-technical creators): enter the Creator Studio “Template Center”, select the corresponding field template (such as “content generation template” “automation script template”), for example, to develop a “template”. “), for example, to develop “Short Video Copy Generation Agent”, based on the “Text Creation Template”, fill in the “input fields (such as video theme, target platform, style) ” “Output Format (e.g. sub-script + title + topic label)”, call the platform’s built-in text generation API, without writing code;
   • Custom development (suitable for technical developers): through the API interface provided by the platform (you need to apply for an API and MCP key in the personal center, which is available to users of the Pro plan and above), develop the Agent logic by yourself – supporting Python/JavaScript and other languages, and integrating your own model or Support Python/JavaScript and other languages, you can integrate your own model or call the big model provided by the platform, and generate a standardized compressed package (including code, configuration files and dependency libraries) through the “Agent Packaging Tool” to ensure compatibility with the MuleRun operating environment after the development is completed.
2. Configure core parameters:
   • Basic information: set the Agent name (must contain core functions, such as “AI hand puppet model generator”), classification labels (such as “Content Creation→3D Modeling”), icons and cover images (must comply with the platform visual specifications and highlight the Agent’s characteristics). );
   • Operation parameters: Define the parameters to be input by the user (e.g. “image upload”, “text description”, “file format selection”), set the parameter constraints (e.g. Define the parameters (such as “image upload”, “text description”, “file format selection”) and set the parameter constraints (such as “image size not more than 10MB”, “text description not less than 20 words”), so as to avoid the failure of the task due to the user’s parameter error;
   • Resource allocation: apply for running resources (such as CPU/memory occupation and concurrency) according to the complexity of the Agent, the basic Agent allocates 1 core and 2G resources by default, and the high load Agent (such as 3D rendering and batch data processing) can apply for capacity expansion after passing the audit.
3. Testing and Optimization:
   • Using the Creator Studio “test environment”, simulate user scenarios, input parameters, run the Agent and check the results, focusing on testing “functional integrity (e.g. whether the results are output as expected), stability (e.g. whether 10 consecutive invocations report errors), efficiency (e.g. whether the time consumed by the task is within a reasonable range)”. Functional integrity (e.g. whether the results are output as expected), stability (e.g. whether 10 consecutive calls report errors), efficiency (e.g. whether the task consumes time within a reasonable range);
   • If there is any “result deviation” (e.g. text generation is not in line with the style) or “running error” in the test, we can locate the problem through the “log analysis tool”, adjust the model parameters or code logic, and test again after optimization to ensure that the results are in line with the expected output. logic, and test again after optimization to ensure that the Agent availability rate is ≥95% (a mandatory requirement for platform audit).

(ii) Submit for audit: Prepare materials according to specifications to improve the passing rate.

1. Fill in the audit application form: Enter the “Submit for Audit” page of Creator Studio and upload the following materials:
   • Basic Information Sheet: Include Agent name, classification, target users, core functions, profit model, analysis of competing products (need to explain the “differences and advantages with the existing Agent”);
   • Demonstration materials: demo video / test result examples, description documents;
   • Technical materials (only custom development needs to provide): code compressed package, list of dependent libraries, API call instructions, to ensure that the code has no security holes (the platform will conduct code scanning to eliminate malicious programs).
2. Waiting for review and modification:
   • Audit cycle: usually 3-5 working days, high-quality creators (e.g. past Agent score ≥ 4.5 stars) can enjoy “Priority Audit Channel”, and the result will be out in 1-2 working days;
   • Audit feedback: If the audit is not passed, the platform will inform the reasons through internal mail (e.g. “Agent function is duplicated with the existing tools”, “demo can’t run normally”, “description document is incomplete”), and the creators need to be informed of the reasons through internal mail. “), creators need to modify according to the feedback within 7 days, and can resubmit for review after modification (maximum 3 resubmission opportunities per quarter, more than need to wait for the next quarter);
   • Passed Audit: The platform will send “Notice of Passed Audit”, informing the Agent of the time to go online (usually within 24 hours after passing the audit) and the initial recommended position (the new Agent will be displayed in the “New Products Zone” for 3 days to get the initial traffic).

(C) Launch: Set up promotion strategy and get initial users.

1. Final Confirmation before Release: Before going online, you need to check the “Parameter Configuration (whether the required items are missing), Pricing (whether it’s reasonable), and Documentation (whether it’s updated to the latest version)”, and then click “Confirm Release” after confirming that there are no errors;
2. Apply for platform support: New Agents can apply for the “Creator Support Program”, which is available to those Agents who meet the conditions (e.g., “first release, functional innovation, clear target users”):
   • Traffic support: official website “new product recommendation” position exposure, platform social media (such as Twitter, Xiaohongshu) promotion, KOL evaluation cooperation;
   • Cash Incentive: When Agent earns ≥100 USD, it can get a minimum of 100 USD and a maximum of 10,000 USD cash incentive (the higher the earnings, the higher the incentive percentage);
3. Self-promotion: Share Agent to personal social accounts (e.g. Discord, GitHub, circle of friends), guide users to try it out, and when new users register and use Agent through the exclusive link, the creator will be rewarded with an additional 500 points (up to a maximum of 100,000 points), and at the same time, the initial user evaluation will be accumulated, which will improve the Agent’s search ranking on the platform.

Post-launch management: from operation optimization to revenue tracking

(A) Data Monitoring and Optimization: Enhance Agent Utilization Rate and Favorable Feedback Rate

1. View operation data: Enter Creator Studio’s “Data Center” and focus on the following indicators:
   • Usage data: daily/weekly usage, user retention rate (e.g., 7-day retention), parameter input error rate (if the error rate is too high, you need to optimize the parameter description or add guidance);
   • Evaluation data: user ratings (1-5 stars), comment content (need to focus on “negative feedback”, such as “slow generation speed” “results do not meet expectations”), complaint rate (need to be (need to control below 1%, otherwise it may affect Agent recommendation);
   • Revenue data: daily points revenue, share amount, user payment conversion rate (if the conversion rate is low, you need to adjust the pricing or optimize the Agent function).
2. Optimize Agent based on feedback:
   • Function optimization: adjust Agent logic based on user comments, for example, “3D Modeling Agent” can upgrade model parameters or add “Detail Enhancement” option if user feedback is “Insufficient model details”; “3D Modeling Agent” can upgrade model parameters or add “Detail Enhancement” option if user feedback is “Insufficient model details”. ” option;
   • Experience Optimization: If user feedback is “complicated parameter input”, reduce non-essential parameters or add “default values” (e.g., select “HD Mode” by default); if “running time consumes too long”, you can optimize the code or apply for platform resource expansion;
   • Version Iteration: Each time a new version is released after optimization, the update content (e.g. “V2.0: new detail enhancement function, generation speed increased by 30%”) needs to be stated in the “version log” to inform users of the value of the update and enhance their willingness to upgrade.

(ii) Earnings Management and Cash Withdrawal: Clear Tracking and Convenient Realization

1. Earnings Viewing and Settlement:
   • The platform splits its revenue with creators on the basis of “daily statistics and weekly settlement” – you can check the real-time revenue of points (100 points equals to about 1 USD) in the “Revenue Center” every day, and the revenue of last week is automatically settled every Monday (after deducting the platform service fee, the basic service fee is 20%). Every Monday, last week’s earnings are automatically settled (after deducting the platform’s service fee, the basic service fee is 20%), and after settlement, the points can be exchanged for US dollars or the platform’s common currency;
   • Support to view “earnings details”, including “user usage records (who uses, number of times used, points consumed)”, “share ratio”, “reward amount (e.g. traffic support reward) Amount (e.g., traffic support rewards)” to ensure that earnings are transparent and traceable.
2. Withdrawal operation:
   • When the account balance ≥ 100 USD, you can apply for cash withdrawal (individual users support PayPal / bank transfer, corporate users need to provide public account information);
   • Enter the “Withdrawal Center”, select the withdrawal method, fill in the receipt information, after submitting the application, the platform will be reviewed and processed within 3-5 working days, and the funds will be transferred to the corresponding account after the review and approval (the handling fee for withdrawals is charged at 1% of the amount withdrawn, and the handling fee can be waived for users of the Pro Plan).

(III) Agent Maintenance and Downgrade: Necessary operations for long-term operation

1. Routine maintenance:
   • Regularly check the Agent operation status (the platform will send “operation abnormality notification”, such as “API call failure” and “insufficient resources”), and timely deal with failures (such as reconfiguring the API key, apply for resource expansion);
   • Pay attention to platform rule updates (e.g., “model interface adjustment” and “content audit specification changes”), if the Agent involves functions or content that do not meet the new rules, it needs to be modified within the specified time to avoid being taken off the shelf.
2. Take the initiative to downgrade or iterate:
   • If the Agent is not used by users for a long time (e.g. ≤10 times in 30 consecutive days) or its functions are outdated, you can take the initiative to choose to “take it down”, and after it is taken down, it can be re-developed and optimized, and submitted for review again;
   • If you need to stop operation, you can apply for “permanent shelving”, before shelving, you need to notify the subscribed users 7 days in advance to protect the rights and interests of users, and the platform will refund the users according to their remaining subscription time.

Fourth, high-quality Agent operation skills: from obtaining traffic to build word of mouth

(i) Optimize search and recommendation ranking

1. Keyword optimization: Add high-frequency search terms (e.g. “AI generation”, “automation”, “3D modeling”) to the Agent name and description, such as “AI Action Figure Model Generator” can be optimized as “AI Action Figure Model Generator – photo to 3D model, no need for modeling skills”, which will enhance the search results in “3D Modeling” and “Photo Generation”. “Photo Generator” to improve search rankings under keywords like “3D Modeling” and “Photo Generation”;
2. Improve user ratings: After the Agent has been used, guide users to rate it (e.g. “If you are satisfied with this experience, please give 5-star positive feedback”), and give a small amount of points to those who give positive feedback (within the scope of the platform’s rules), and Agents with ratings of ≥4.5 stars will enter the “Quality Agent Recommendation Pool”. Agent with rating ≥4.5 stars will enter the “Quality Agent Recommendation Pool” and get more exposure on the homepage.

(ii) Create personal brands for creators

1. Improve the creator’s homepage: upload a professional avatar and profile (e.g. “Focusing on 3D modeling AI tool development, serving 100,000 users cumulatively”), display past quality Agents and user evaluations, and enhance user trust;
2. Interactive operation: respond to user questions (such as “how to adjust model details”) and collect user needs (such as “I hope to support more model styles”) in the Agent comment area in a timely manner, so that users can feel the “continuous maintenance ” attitude to enhance the repurchase rate;
3. Participate in platform activities: actively participate in the “Agent Creation Contest” “Vertical Field Thematic Activities” organized by MuleRun (e.g. “E-commerce Tools Theme” “Game Aid”), the winning Agent can obtain exclusive promotion resources and quickly increase its popularity.

(iii) Utilize platform ecological resources

1. Join the creator community: Join MuleRun’s official Discord creator channel ( https://discord.com/invite/8Ccd2ksqx6), exchange development experience with other creators, and obtain the platform’s latest policies and resources (e.g., “Qualification for internal testing of new model interfaces”). (e.g. “new model interface qualification for internal testing”);
2. Apply for “Quality Creator” Certification: When Agent has been used ≥1000 times, rated ≥4.5 stars, and has no violation records, it can apply for “Quality Creator” certification, and after passing the certification, it can obtain “higher After passing the certification, you can get “higher share ratio, exclusive customer service, priority review” and other rights and benefits, which will further enhance the revenue and operational efficiency.

Summarize: The key logic from “skills” to “sustainable income”.