Introduction

ApiGo is a general-purpose API development and management platform built on a microservice architecture + dynamic data sources + dynamic standard SQL. It quickly wraps enterprise databases, third-party HTTP services and existing APIs into governable, exposable and subscribable REST APIs and MCP tools.

The platform separates "API definition, runtime governance and API execution" into independent services: the admin side handles configuration and publishing, the gateway handles unified governance, and the execution side performs the actual execution of SQL, proxy and orchestration. Configuration is persisted in MySQL and distributed through Redis, so changes take effect without a restart.

Core capabilities

Capability	Description
Four development modes	SQL script, single table (zero code), proxy forwarding and API orchestration, covering database queries, third-party API wrapping and multi-API aggregation
Multiple data sources	11 SQL data sources + 5 NoSQL / new-type data sources, with unified metadata browsing and read-only queries
Standard SQL + dynamic tags	Supports MyBatis-style if, foreach, where, set and choose tags as well as #{} prepared parameters
Unified gateway governance	Authentication, credential authorization, subscription quota, rate limiting and circuit breaking, caching, IP allow/deny lists, WAF, Mock and logging are all completed at the gateway
Open platform and subscriptions	API Hub listing, online documentation, online testing, online signing, subscription application and approval, daily call quota and expiration reminders
Native MCP	APIs can be published as MCP Tools; both calling credentials and agents can connect to AI clients as MCP Servers
AI assistant	Natural language data queries, SQL generation and read-only execution, tool timeline, clarification cards and write-operation confirmation, with the whole process recorded
Multi-project isolation	Project spaces isolate APIs, data sources, groups, environments and policies; cross-project data source authorization is supported
Security and audit	JWT, SSO, RBAC, AppKey/AppSecret signatures, SM3/HmacSHA256, operation logs, call logs and alerting
License	Offline activation for jar, Docker and Kubernetes deployment forms

Supported data sources

Category	Data sources
SQL (11)	MySQL, Oracle, SQLServer, PostgreSQL, TiDB, OpenGauss, OceanBase, DM (dm), Kingbase (kingbase8), TDSQL, StarRocks
NoSQL / new-type (5)	ElasticSearch, TDengine, ClickHouse, DolphinDB, Hive

Syntax differences

Non-relational data sources such as ElasticSearch and Hive do not support the full standard SQL, nor dynamic tags such as if; the platform performs common syntax conversions at the driver layer. When table or field names contain special characters such as -, wrap them in double quotes.

Product architecture at a glance

Typical use cases

1. Controlled AI access to enterprise data: wrap order, customer, inventory and other queries as MCP Tools; agents discover tools through tools/list and call them through tools/call, while the platform retains authentication, rate limiting, quota and complete logs.
2. Enterprise API open portal: developers browse public APIs in API Hub, view documentation and debug online, submit subscription applications and, once approved, access APIs with calling credentials.
3. Service-izing existing systems: without modifying legacy systems, quickly wrap database tables and third-party APIs into standard APIs through SQL or proxy modes.
4. Multi-API aggregation: use orchestration mode to merge, filter, compute, map and sort the results of multiple published APIs, exposing only one API externally.
5. Unified traffic governance: configure rate limiting, circuit breaking, caching, IP allow/deny lists, WAF and alerting at the gateway to protect backend data sources.

Product boundaries

• The platform focuses on API building, governance and exposure and does not undertake business logic development beyond business transaction orchestration; complex business logic should still live in business systems.
• SQL capabilities for non-relational data sources depend on driver conversion and are not guaranteed to cover the full standard SQL syntax.
• Open APIs return JSON by default; proxy APIs can be configured to pass through binary raw responses.
• Write operations of the AI assistant require manual confirmation, and the tool scope is constrained by the current user's menu permissions.
• MCP calls reuse the filter chain of regular APIs, but API caching does not apply to MCP calls.

Next steps

• Quick start: complete the full flow from login to your first call.
• Core concepts: first understand the relationships among projects, environments, versions, credentials and subscriptions.
• Architecture overview: learn about the architecture that separates design time from runtime.