Overview
Makers Agents is an out-of-the-box Agent development platform. Developers can focus on business logic and, upon deployment, gain access to the full capabilities of Agent runtime, sandbox tools, conversation memory, locally debuggable end-to-end tracing, and built-in models.
Platform Advantages
Out-of-the-box: The tools, memory, observability, and models required for Agent operation are all automatically activated upon deployment, requiring no installation or integration.
Flexible and Open: Choose any framework (Claude / OpenAI / LangGraph / CrewAI), with no language restrictions (JS / Python).
Seamless Integration: The Web and Agent share the same project, and accounts, deployments, monitoring, and domains are centrally managed.
Core Capabilities
Capability | Description |
Managed Runtime | Hosts LLM invocations, Agent loop orchestration, and business logic, routes by session, and automatically scales. |
Sandbox Tool | Provides two layers of independent yet interconnected APIs for LLMs and developers, with browser, code execution, Shell, and file operations all running in an isolated sandbox environment. |
Conversation Storage | Adapts to various Agent frameworks to provide memory management capabilities and offers a universal API for managing sessions and messages. |
Observability | Automatically collects call traces, performs zero-intrusion instrumentation, and provides a unified view of trace data on local / cloud panels. |
Built-in Models | Automatically injects model keys and provides a limited-time monthly free model quota. |
Running mode
The platform provides two types of Runtime, distinguished by directory for their intended use. Within the same project, they can be combined in any configuration.
Directory | Runtime Mode | Purpose | Injected context |
agents/ | Session mode (routes to the same instance based on conversation_id for stickiness and reuses memory state). | LLM invocation, Agent loop, and long-running task orchestration | request / env / store / tools / sandbox / tracer |
cloud-functions/ | Request mode (stateless, one execution environment per request, with elastic scaling). | Non-LLM business logic: data queries, auxiliary APIs, and so on | request / env / agent.store |
agents/ is designed specifically at the runtime level to handle genuine Agent workloads:Session Stickiness: Requests with the same
conversation_id are routed to the same instance, reusing the context, cache, and connections in its memory while the conversation is active.Extended Execution Duration: A single execution defaults to 5 minutes and can be configured for up to 1 hour, which is sufficient to support long-running tasks such as multi-turn Agent loops, repeated tool invocations, and in-depth research.
cloud-functions/ is a lightweight, stateless, and elastically scalable per request mode: each request is reclaimed immediately after use, incurring no persistent cost, making it suitable for short, straightforward logic such as data queries and auxiliary APIs.Project Structure
project/agents/ ← Conversational Mode Runtime (LLM / Agent Loop)customer-service/index.ts → POST /customer-servicestop.ts → POST /customer-service/stopcloud-functions/ ← Request Mode Runtime (Non-LLM Business)conversations/list.ts → GET /conversations/list (Reads the store)edgeone.json ← Project-level configuration
Scenarios
Conversational Assistant: It completes continuous tasks in multi-turn interaction scenarios such as knowledge base Q&A and shopping order placement by leveraging memory and tool invocation.
File Processing: It parses CSV files, processes multimodal files such as PDFs and images, and reads and analyzes file content within a sandbox.
Scheduled Task: It automatically collects data and generates structured reports on a scheduled basis for periodic tasks such as trending hot news crawling and aggregation.
Content Generation: In scenarios such as blog posts with images and marketing campaign planning, long-form articles and accompanying images are collaboratively produced by multiple roles or through multiple steps.
Process Orchestration: It manages stateful and branching processes such as deep research and email handling, based on node transition states, validation, and manual review.
Programming Assistant: In scenarios such as AI website builders, it modifies code and previews the execution results within a cloud sandbox.
Built-in Models
New users receive a limited-time trial quota of 500,000 Tokens, which is calculated per account and shared across all projects. When deployment is performed using a template, the following environment variables are automatically injected, and your business code can read them via
context.env.Variable | Description |
AI_GATEWAY_API_KEY | Gateway authentication Key, compatible with the OpenAI protocol |
AI_GATEWAY_BASE_URL | Gateway address, default https://ai-gateway.edgeone.link/v1 |
Any SDK compatible with the OpenAI protocol can be directly integrated. To switch models, you only need to modify the
model field without changing the Key. For details, refer to the Models Overview.Free Edition Quota Limits
EdgeOne Makers currently offers a free edition. All quotas listed below are those currently in effect for the free edition. The commercial edition is being planned. Specific pricing and quotas will be updated via the console announcement after they are released. Quotas for the free edition are enforced more leniently before the commercial edition officially launches. Even if usage exceeds limits, your business stability will be prioritized. If your quota is insufficient, you can submit a ticket to request an increase. The platform will continuously monitor for abusive behavior.
Agents
Item | Free Edition |
Number of Executions | 200,000 /month |
Total Memory Time | 100,000 GB-s/month |
Maximum runtime per request | 1800 seconds |
Maximum session idle time | 300 seconds |
Maximum concurrent running sessions | 40 |
Sandbox
Item | Free Edition |
Total Memory Time | 100,000 GB-s/month |
Maximum concurrent running sessions | 20 |
Maximum runtime per instance | 3600 seconds |
Default timeout | 300 seconds |
Core Concepts at a Glance
Concept | Description |
context | A unified context object injected by the platform on each request, carrying all sub-capabilities listed below. |
context.request | A standard Web Request object (including body / headers / signal) for reading input parameters and listening for interrupts. |
context.env | An access point for environment variables and secrets, equivalent to the secure reading of process.env / os.environ |
context.store | A conversation-level dialogue storage that is automatically associated by conversation_id and supports native message objects from five major frameworks. |
context.tools | A tool list for LLM, automatically packaged into a form that can be directly consumed by the target framework based on the framework field. |
context.sandbox | Sandbox atomic APIs ( commands / files / browser / runCode) for running commands, reading/writing files, operating browsers, and executing code in an isolated environment. |
context.tracer | An OpenTelemetry-style manual instrumentation entry point that shares the same trace link as the platform's automatic collection. |
context.agent.store | Same as context.store and exposed in Cloud Functions. |
conversation_id | A unique identifier for a session, serving as the key for routing, Store association, and Sandbox instance association. |
run_id | The unique identifier for an execution, around which traces, logs, and stop operations are organized. |
framework | A field in edgeone.json that determines the framework adaptation form of context.tools / context.store |