OceanBase MCP
Provides Docker images and Docker Compose templates for deploying the OceanBase MCP server in containerized environments.
Provides CI/CD workflows using GitHub Actions for building, testing, and releasing the MCP server.
Provides Kubernetes deployment templates for running the OceanBase MCP server in a Kubernetes cluster.
Provides an NGINX reverse proxy configuration example for securing the HTTP MCP server.
Exposes a Prometheus metrics endpoint for monitoring the MCP server's performance and health.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@OceanBase MCP list tables in schema app_schema"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
OceanBase MCP
OceanBase MCP is a strictly read-only MCP server for OceanBase. It lets OpenCode and other MCP clients inspect schemas, metadata, query plans, ER relationships, and limited sample data without exposing write capability.
OceanBase MCP 是面向 OceanBase 的严格只读 MCP server。OpenCode 和其他 MCP 客户端可以通过它查看 schema、元数据、执行计划、ER 关系和受限样例数据,同时不暴露写入能力。
What It Provides / 能力概览
Area | 中文 | English |
Database modes | 支持 OceanBase MySQL 模式和 Oracle 模式。 | Supports OceanBase MySQL and Oracle compatibility modes. |
Query workflow | 元数据优先、查询预览、分页查询、结果摘要和错误提示。 | Metadata-first query workflow with preview, paging, summaries, and error hints. |
Metadata | 表结构、DDL、索引、约束、分区、统计信息和 Oracle 对象元数据。 | Table structure, DDL, indexes, constraints, partitions, statistics, and Oracle object metadata. |
ER diagrams | Mermaid、JSON、自包含 HTML 和本地分片导出。 | Mermaid, JSON, self-contained HTML, and local chunked exports. |
SQL assistance | 只读 SQL 生成、改写、lint、执行计划分析和性能建议。 | Read-only SQL generation, rewrite, lint, plan analysis, and performance advice. |
Change review | 元数据快照、差异对比、项目 SQL 使用索引和影响报告。 | Metadata snapshots, diffs, project SQL usage index, and impact reports. |
Deployment | stdio、本地 OpenCode、HTTP、Docker、Kubernetes、systemd 和 Nginx 示例。 | stdio, local OpenCode, HTTP, Docker, Kubernetes, systemd, and Nginx examples. |
Complete tool list / 完整工具列表按首次引入版本排序:
Related MCP server: db-connect-mcp
Read-Only Contract / 只读边界
The server is read-only by design:
No MCP tool can write data or change database objects.
OB_MCP_READ_ONLY=trueis the default and cannot be disabled.OB_MCP_READ_ONLY=falsemakes the server refuse to start.User SQL is limited to
SELECT,WITH, and read-onlyEXPLAIN.SELECT *, multi-statement SQL, write/control keywords, executable comments, filesystem functions, sleeps, locks, and sequenceNEXTVALare blocked.User SQL runs in a database read-only transaction by default.
服务从设计上只读:
没有任何 MCP 工具可以写入数据或修改数据库对象。
OB_MCP_READ_ONLY=true是默认值,且不能关闭。设置
OB_MCP_READ_ONLY=false会导致服务拒绝启动。用户 SQL 只允许
SELECT、WITH和只读EXPLAIN。默认拦截
SELECT *、多语句、写入/控制关键字、可执行注释、文件函数、sleep、lock 和序列NEXTVAL。用户 SQL 默认在数据库只读事务中执行。
Even if the configured database account has write grants, the MCP surface remains read-only. A SELECT-only database account is still recommended as defense in depth.
即使配置的数据库账号具备写权限,MCP 对外仍然只能读。仍建议使用只有 SELECT 和元数据查看权限的数据库账号作为纵深防御。
Quick Start / 快速开始
Node.js 20 or newer is required.
需要 Node.js 20 或更高版本。
npm install -g @sevoniva/oceanbase-mcp@1.5.0
ob-mcp init --mode mysql --schema app_schema
ob-mcp doctorUse npx when you do not want a global install:
不想全局安装时,可以直接用 npx:
npx --registry=https://registry.npmjs.org/ --yes --package=@sevoniva/oceanbase-mcp@1.5.0 ob-mcp --helpBuild from source:
从源码构建:
git clone https://github.com/sevoniva/oceanbase-mcp.git
cd oceanbase-mcp
npm ci
npm run build
cp .env.example .env
npm run config:doctor
npm run startOpenCode / 接入 OpenCode
For local OpenCode use, generate a project config and then verify the MCP connection:
本地使用 OpenCode 时,先生成项目配置,再检查 MCP 连接:
ob-mcp init --mode mysql --schema app_schema
opencode mcp listAn npx-based OpenCode entry looks like this:
基于 npx 的 OpenCode 配置示例:
{
"mcp": {
"oceanbase": {
"type": "local",
"command": "npx",
"args": [
"--yes",
"--package=@sevoniva/oceanbase-mcp@1.5.0",
"ob-mcp"
],
"enabled": true,
"environment": {
"OB_CONFIG_PATH": ".env"
}
}
}
}For the recommended OpenCode query workflow, see:
推荐的 OpenCode 查询流程见:
Common Workflows / 常用流程
Workflow | 中文 | English |
Query data safely | 先搜索表和字段,再预览 SQL,最后分页查询。 | Search tables and columns, preview SQL, then run a paged query. |
Understand a table | 获取表上下文、字段画像、索引、约束、统计信息和相关表。 | Get table context, column profiles, indexes, constraints, statistics, and related tables. |
Draw ER diagrams | 小范围用 MCP 返回 Mermaid/HTML,大 schema 用 CLI 分片导出。 | Use MCP Mermaid/HTML for small scopes and CLI chunked export for large schemas. |
Review schema changes | 迭代前后采集元数据快照,生成 diff 和影响报告。 | Capture metadata snapshots before and after an iteration, then generate diff and impact reports. |
Review project SQL | 扫描本地 SQL 片段,结合元数据校验表字段和影响面。 | Scan local SQL snippets and validate tables, columns, and impact with metadata. |
Documentation / 文档
Topic | 中文 | English |
Start here | ||
Tools | ||
OpenCode | ||
CLI | ||
Configuration | ||
SQL assistance | ||
ER diagrams | ||
Data dictionary | ||
Metadata change | ||
Deployment | ||
Architecture | ||
Security | ||
Examples |
Release Timeline / 版本路线
Version | 中文 | English |
v0.3.0 | 建立只读 MCP 网关、OceanBase MySQL/Oracle 连接、基础 SQL 策略、表元数据和脱敏样例数据。 | Read-only MCP gateway, OceanBase MySQL/Oracle connectivity, baseline SQL policy, table metadata, and masked sample rows. |
v0.4.0 | 增加 HTTP 部署、安全加固、Host/IP allowlist、限流、健康检查、指标、审计日志和数据库权限核验。 | HTTP deployment, security hardening, Host/IP allowlists, rate limits, health checks, metrics, audit logs, and database permission checks. |
v0.5.0 | 补齐开发元数据:表 DDL、统计信息、Oracle 对象、关系发现、ER 图、执行计划分析和性能建议。 | Developer metadata: table DDL, statistics, Oracle objects, relationship discovery, ER diagrams, plan analysis, and performance advice. |
v0.5.1 | 改善 OpenCode 查询流程:表上下文、查询预览、分页查询、结果结构、错误提示、字段画像和业务字典。 | OpenCode query workflow: table context, query preview, paged queries, result structure, error hints, column profiling, and business glossary. |
v1.0.0 | 完成本地开发助手能力:初始化/诊断、统一搜索、SQL 辅助、查询会话、数据理解、ER HTML、文档生成、项目 SQL 扫描和离线交付。 | Local developer assistant scope: init/doctor, unified search, SQL assistance, query sessions, data understanding, ER HTML, documentation, project SQL scanning, and offline delivery. |
v1.1.0 | 增加元数据快照、快照对比、SQL 使用索引、影响分析、变更报告和关系路径解释,用于迭代前后评审。 | Metadata snapshots, snapshot diff, SQL usage index, impact analysis, change reports, and relationship paths for iteration review. |
v1.1.1 | 修复大 schema ER 生成稳定性,增加批量元数据加载、ER HTML 浏览器优化和本地分片导出规划。 | Large-schema ER stability, batched metadata loading, improved ER HTML browser, and chunked local export planning. |
v1.1.2 | 修复本地 OpenCode 连续查询稳定性,放宽本地 HTTP 默认限流,增加故障日志,并兼容 Oracle 常见命名占位符。 | Local OpenCode query stability, higher local HTTP rate-limit default, crash/slow-call logging, and Oracle named bind compatibility. |
v1.1.3 | 优化大 schema ER 规划为轻量估算,增加分片策略参数,并完成 200+ 表本地验证。 | Lightweight large-schema ER planning, chunk strategy options, and 200+ table local validation. |
v1.1.4 | 减少 ER 默认关系发现的重复元数据读取,并对大范围 inline ER HTML 提前降级。 | Reduced duplicate ER metadata reads and early fallback for large inline ER HTML. |
v1.1.5 | 增加 npmjs 分发,包名为 | npmjs distribution under |
v1.1.6 | 修正 npmjs 启动示例,补充 Apache-2.0 包许可元数据和 Trusted Publishing 发布流程,并验证公开 npm 安装。 | Correct npmjs startup examples, add Apache-2.0 package license metadata, document Trusted Publishing, and verify public npm installation. |
v1.2.0 | 重构 ER 元数据地图:稳定图 ID、交互式 graph 画布、cluster 目录导出、跨 cluster 关系索引、诊断元数据和响应大小日志。 | Rebuilt ER metadata maps with stable graph IDs, interactive graph canvas, cluster directory export, cross-cluster relationship index, diagnostic metadata, and response-size logging. |
v1.3.0 | 强化稳定性和可维护性:DB 调用预算、工具截止时间、结构化日志、崩溃回归测试、DB 运行时模块拆分和大库验证记录。 | Stability and maintainability: DB call budgets, tool deadlines, structured logs, crash regression coverage, DB runtime module split, and large-schema validation records. |
v1.4.0 | 精简 npm 包内容,移除运行时不需要的历史 release 文档和 smoke 脚本,保留源码归档用于完整构建。 | Trimmed npm package contents by excluding release-note history and smoke scripts that are not needed at runtime; source archives remain complete for source builds. |
v1.5.0 | 重整公开文档入口、拆分工具清单、清理发布文案,并增加公开文档措辞检查。 | Public documentation refresh, separate tool list, cleaner release wording, and public documentation wording checks. |
Distribution / 分发
The public npm package is @sevoniva/oceanbase-mcp, and the CLI commands are ob-mcp, ob-mcp-http, and ob-mcp-doctor. Release assets also include a source archive, CycloneDX SBOM, and amd64/arm64 offline image archives.
公开 npm 包名为 @sevoniva/oceanbase-mcp,CLI 命令为 ob-mcp、ob-mcp-http 和 ob-mcp-doctor。Release 附件同时提供源码归档、CycloneDX SBOM,以及 amd64/arm64 离线镜像归档。
npm install -g @sevoniva/oceanbase-mcp@1.5.0
ob-mcp --helpSource archive build:
源码归档构建:
tar -xzf oceanbase-mcp-1.5.0-source.tar.gz
cd oceanbase-mcp-1.5.0
npm ci
npm run build
npm run startThe source archive excludes node_modules and dist. Your npm registry mirror must provide the dependencies locked by package-lock.json.
源码归档不包含 node_modules 和 dist。npm registry 镜像需要能提供 package-lock.json 中锁定的依赖。
Validation / 验证
npm test
npm run docs:check
npm audit --omit=dev --audit-level=moderate
npm pack --dry-run
docker build -t oceanbase-mcp:local .Smoke tests are documented in Deployment and 部署指南.
Available Tools
85 toolsob_analyze_change_sql_impactAnalyze Change SQL ImpactC
Analyze which indexed project SQL snippets are impacted by a metadata snapshot diff.
Inputs: sqlIndex (required), metadataDiff (required), profile, maxResults.
ob_analyze_metadata_impactAnalyze Metadata ImpactC
Summarize metadata diff impact using optional SQL usage index and business glossary evidence.
Inputs: metadataDiff (required), profile, sqlIndex, maxResults, +1 more.
ob_analyze_query_planAnalyze Query PlanC
Run EXPLAIN for a strictly read-only SQL statement and return evidence-based plan findings.
Inputs: sql (required), params, maxRows, profile.
ob_build_selectBuild SELECTA
Build a parameterized read-only SELECT from validated table, columns, filters, ordering, and pagination without executing it.
Inputs: table (required), page, schema, columns, +4 more.
ob_capture_metadata_snapshotCapture Metadata SnapshotB
Capture a bounded metadata-only OceanBase snapshot for later offline diff and impact analysis.
Inputs: schema, profile, maxTables, maxObjects, +6 more.
ob_check_data_qualityCheck Data QualityB
Run bounded read-only data quality checks for nulls, duplicate key risk, dimensions, and time columns.
Inputs: table (required), schema, columns, profile, +3 more.
ob_check_opencode_setupCheck OpenCode SetupC
Check local OpenCode config, OceanBase agent, and query skill files for the current project.
Inputs: profile.
ob_close_query_sessionClose Query SessionB
Close an in-memory query session for the selected profile.
Inputs: sessionId (required), profile.
ob_compare_metadata_snapshotsCompare Metadata SnapshotsB
Compare two metadata-only snapshots and return evidence-based structural changes.
Inputs: afterSnapshot (required), beforeSnapshot (required), profile, riskOnly, +2 more.
ob_compare_query_resultsCompare Query ResultsC
Compare two already returned query results without reading the database or storing row values.
Inputs: leftRows (required), rightRows (required), columns, profile, +5 more.
ob_compare_schemasCompare SchemasB
Compare two visible schemas by table list and bounded per-table metadata differences.
Inputs: leftSchema (required), rightSchema (required), profile, maxTables, +2 more.
ob_compare_tablesCompare TablesC
Compare two visible tables by columns, indexes, constraints, and partitions.
Inputs: leftTable (required), rightTable (required), profile, leftSchema, +1 more.
ob_config_diagnosticsRun Config DiagnosticsB
Check profile, policy, TLS, HTTP, and production hardening settings.
Inputs: no named parameters.
ob_convert_sql_dialectConvert SQL DialectC
Convert read-only SQL between OceanBase MySQL and Oracle modes with documented limitations.
Inputs: sql (required), targetMode (required), profile, sourceMode.
ob_describe_tableDescribe TableC
Return column metadata for an allowed OceanBase table.
Inputs: table (required), schema, profile, maxColumns.
ob_detect_dimension_columnsDetect Dimension ColumnsC
Detect likely status, type, channel, region, flag, and category columns from visible table metadata.
Inputs: table (required), schema, profile, maxColumns.
ob_detect_time_columnsDetect Time ColumnsC
Detect likely create, update, and business time columns from visible table metadata.
Inputs: table (required), schema, profile, maxColumns.
ob_explainExplain QueryC
Run EXPLAIN for a strictly read-only SQL statement.
Inputs: sql (required), params, maxRows, profile.
ob_explain_errorExplain SQL ErrorC
Explain a SQL error with policy context and visible metadata hints without executing the SQL.
Inputs: sql (required), errorMessage (required), limit, schema, +1 more.
ob_explain_project_queryExplain Project QueryC
Explain a read-only SQL snippet from text or a local project file without executing it.
Inputs: sql, line, params, schema, +4 more.
ob_explain_relationshipExplain Table RelationshipB
Explain visible relationship evidence between two OceanBase tables without inventing missing relationships.
Inputs: sourceTable (required), targetTable (required), depth, schema, +5 more.
ob_export_result_previewExport Result PreviewA
Generate a bounded Markdown or CSV preview string from already returned rows without writing files.
Inputs: rows (required), format, columns, maxRows, +1 more.
ob_find_columnsFind ColumnsB
Find visible columns across tables by name, comment, data type, table pattern, or business term.
Inputs: limit, query, schema, profile, +3 more.
ob_find_recent_rowsFind Recent RowsC
Read recent rows from an allowed table using an explicit or metadata-detected time column with safe pagination.
Inputs: table (required), page, schema, columns, +3 more.
ob_find_sql_usagesFind SQL UsagesB
Find project SQL usages from a static SQL usage index without executing SQL or reading project files.
Inputs: sqlIndex (required), table, column, schema, +3 more.
ob_find_tables_by_business_termFind Tables by Business TermB
Find visible candidate tables from a business term using glossary, catalog metadata, comments, and relationship evidence.
Inputs: term (required), limit, schema, profile, +2 more.
ob_find_unknown_columns_in_sqlFind Unknown Columns In SQLA
Find unknown tables or columns in SQL and return visible metadata-based candidates without executing the SQL.
Inputs: sql, line, schema, profile, +2 more.
ob_generate_data_dictionaryGenerate Data DictionaryC
Generate bounded JSON, Markdown, or simple HTML data dictionary output from visible metadata.
Inputs: title, format, schema, profile, +3 more.
ob_generate_entity_diagramGenerate Entity DiagramC
Generate a local entity-focused ER diagram from a business term or table, with Mermaid, JSON, and optional HTML.
Inputs: depth, table, title, schema, +9 more.
ob_generate_er_diagramGenerate ER DiagramC
Generate Mermaid and JSON ER metadata from allowed OceanBase table metadata and relationships.
Inputs: depth, table, schema, tables, +7 more.
ob_generate_er_documentGenerate ER DocumentationC
Generate Markdown, PlantUML, and DOT documentation from allowed ER metadata and relationship evidence.
Inputs: depth, table, title, schema, +9 more.
ob_generate_er_htmlGenerate ER HTMLC
Generate a self-contained HTML ER diagram from allowed OceanBase metadata and relationship evidence.
Inputs: depth, table, title, format, +9 more.
ob_generate_mapper_contextGenerate Mapper ContextC
Generate read-only table and column context for a MyBatis mapper, SQL file, or explicit table.
Inputs: sql, line, table, format, +5 more.
ob_generate_metadata_change_reportGenerate Metadata Change ReportC
Generate a Markdown or simple HTML metadata change report from diff and impact evidence.
Inputs: metadataDiff (required), title, format, impact, +2 more.
ob_generate_onboarding_docGenerate Onboarding DocumentationC
Generate bounded onboarding documentation from visible OceanBase metadata and optional business glossary hints.
Inputs: title, format, schema, profile, +6 more.
ob_generate_performance_reportGenerate Performance ReportB
Generate a read-only SQL performance advisory report from EXPLAIN, table statistics, and index metadata.
Inputs: sql (required), params, maxRows, profile.
ob_generate_queryGenerate Read-only Query TemplatesB
Generate read-only SQL templates from policy-filtered table metadata without executing them.
Inputs: table (required), schema, maxRows, profile, +2 more.
ob_generate_query_docGenerate Query DocumentationB
Generate bounded JSON or Markdown documentation for a SQL snippet without executing the target query.
Inputs: sql, line, title, format, +6 more.
ob_generate_query_examplesGenerate Query ExamplesB
Generate scenario-based read-only SQL examples from policy-filtered table metadata without executing them.
Inputs: table (required), schema, maxRows, profile, +2 more.
ob_generate_schema_docGenerate Schema DocumentationC
Generate bounded JSON or Markdown documentation for visible tables in one schema.
Inputs: title, format, schema, profile, +5 more.
ob_generate_table_docGenerate Table DocumentationC
Generate bounded JSON or Markdown documentation for one visible table without returning sample values.
Inputs: table (required), title, format, schema, +6 more.
ob_get_cache_statusGet Metadata Cache StatusB
Return profile-scoped metadata cache settings and counters without exposing cached content.
Inputs: profile.
ob_get_constraintsGet Table ConstraintsC
Return constraints for an allowed OceanBase table.
Inputs: table (required), schema, profile.
ob_get_database_dictionaryGet Database DictionaryC
Generate a bounded JSON or Markdown data dictionary from visible table and column metadata.
Inputs: format, schema, profile, maxTables, +2 more.
ob_get_indexesGet Table IndexesC
Return indexes for an allowed OceanBase table.
Inputs: table (required), schema, profile.
ob_get_object_metadataGet Oracle Object MetadataC
Return bounded, policy-filtered metadata for one Oracle view, sequence, synonym, procedure, function, or package.
Inputs: objectName (required), objectType (required), schema, profile.
ob_get_partitionsGet Table PartitionsC
Return partition metadata for an allowed OceanBase table.
Inputs: table (required), schema, profile.
ob_get_query_sessionGet Query SessionB
Return a read-only query session context without exposing stored parameter values.
Inputs: sessionId (required), profile.
ob_get_schema_mapGet Schema MapC
Return a compact, policy-filtered schema map with table summaries, key fields, and relationship hints.
Inputs: schema, profile, maxTables, tablePattern, +4 more.
ob_get_table_contextGet Table ContextC
Return the common read-only table context needed for SQL drafting and data inspection.
Inputs: table (required), schema, profile, sampleLimit, +5 more.
ob_get_table_ddlGet Table DDLC
Return native or structured DDL metadata for an allowed OceanBase table.
Inputs: table (required), schema, profile, maxColumns.
ob_get_table_schemaGet Table SchemaC
Return policy-filtered table columns and types for one table.
Inputs: table (required), schema, profile, maxColumns.
ob_get_table_statsGet Table StatisticsB
Return table-level optimizer statistics and metadata counts for an allowed OceanBase table.
Inputs: table (required), schema, profile.
ob_get_table_summaryGet Table SummaryC
Return a compact table summary with inferred purpose, key fields, relationship hints, and query entry points.
Inputs: table (required), schema, profile, maxColumns, +2 more.
ob_get_usage_guideGet OceanBase MCP Usage GuideA
Return the recommended read-only OpenCode workflow, safety rules, and MySQL/Oracle notes.
Inputs: profile.
ob_get_value_distributionGet Value DistributionB
Return bounded top-value distribution for one allowed column, with masking and sample limits enforced.
Inputs: table (required), column (required), topN, schema, +2 more.
ob_lint_sqlLint SQLB
Lint read-only SQL for policy, dialect, pagination, ordering, sensitive fields, and optional EXPLAIN risk without returning data.
Inputs: sql (required), params, maxRows, profile, +1 more.
ob_list_capabilitiesList OceanBase MCP CapabilitiesC
List supported tool groups, tool versions, active compatibility mode, and read-only limits.
Inputs: profile.
ob_list_objectsList Oracle ObjectsC
List policy-filtered Oracle object metadata such as views, sequences, synonyms, procedures, functions, and packages.
Inputs: limit, query, schema, profile, +1 more.
ob_list_profilesList OceanBase ProfilesA
List configured OceanBase MCP profiles with redacted settings.
Inputs: no named parameters.
ob_list_schemasList SchemasC
List schemas/databases visible to the current OceanBase user, filtered by access policy.
Inputs: profile.
ob_list_tablesList TablesB
List tables visible to the current OceanBase user, filtered by access policy.
Inputs: limit, schema, profile.
ob_pingPing OceanBaseC
Verify the configured OceanBase connection.
Inputs: profile.
ob_plan_er_exportPlan ER ExportC
Plan local ER HTML export commands for large OceanBase schemas without returning large inline HTML.
Inputs: depth, table, format, schema, +12 more.
ob_policy_explainExplain SQL Policy DecisionB
Explain whether a SQL statement is allowed and which read-only or access-policy checks apply.
Inputs: sql (required), params, profile.
ob_prepare_queryPrepare Read-only QueryC
Prepare candidate tables, fields, read-only SQL drafts, and safety previews from visible metadata without executing the target query.
Inputs: question (required), table, schema, maxRows, +4 more.
ob_profile_columnsProfile Table ColumnsC
Return bounded, policy-filtered column profiles for an allowed OceanBase table.
Inputs: table (required), schema, columns, profile, +4 more.
ob_profile_tableProfile TableC
Return a bounded table-level profile with metadata counts, key fields, time/status fields, and sampled null overview.
Inputs: table (required), schema, profile, rowLimit, +4 more.
ob_queryRun Strict Read-Only QueryC
Run a strictly read-only SQL statement against OceanBase.
Inputs: sql (required), params, maxRows, profile.
ob_query_next_pageQuery Next Session PageC
Run the next page for a read-only query session without bypassing row, timeout, output, or policy limits.
Inputs: sessionId (required), orderBy, profile, pageSize.
ob_query_pageRun Paged Read-Only QueryB
Run a strictly read-only SELECT/WITH query with server-side pagination.
Inputs: sql (required), page, params, orderBy, +2 more.
ob_query_previewPreview Read-only QueryA
Preview policy, sensitive-column, and EXPLAIN risk for a read-only SQL statement without executing the target query.
Inputs: sql (required), params, maxRows, profile.
ob_query_prev_pageQuery Previous Session PageA
Run the previous page for a read-only query session without bypassing row, timeout, output, or policy limits.
Inputs: sessionId (required), orderBy, profile, pageSize.
ob_query_refineRefine Query Session SQLB
Build a refined read-only SELECT from a query session using structured columns, filters, ordering, and pagination without executing it.
Inputs: sessionId (required), page, columns, filters, +6 more.
ob_refresh_metadata_cacheRefresh Metadata CacheB
Clear in-memory metadata cache entries for the active profile, optionally scoped by schema or table.
Inputs: table, schema, profile.
ob_rewrite_sqlRewrite SQLB
Rewrite SELECT/WITH SQL into a safer read-only form without executing it.
Inputs: sql (required), params, maxRows, orderBy, +2 more.
ob_sample_rowsSample RowsB
Return a small, policy-filtered sample from an allowed table. Sensitive-looking fields are always masked.
Inputs: table (required), limit, schema, profile.
ob_search_business_glossarySearch Business GlossaryB
Search optional business terms mapped to allowed OceanBase schemas, tables, and columns.
Inputs: query (required), limit, schema, profile.
ob_search_catalogSearch CatalogC
Search schemas, tables, columns, and comments, filtered by access policy.
Inputs: query (required), limit, schema, profile.
ob_search_everythingSearch OceanBase MetadataC
Search visible schemas, tables, columns, comments, and business glossary entries with confidence and evidence.
Inputs: query (required), limit, schema, profile, +1 more.
ob_start_query_sessionStart Query SessionB
Create an in-memory read-only query context for follow-up paging, refinement, and summarization.
Inputs: sql, page, table, params, +7 more.
ob_summarize_resultSummarize Query ResultC
Summarize already returned query rows without reading the database.
Inputs: rows (required), columns, profile, sessionId, +1 more.
ob_validate_sql_snippetValidate SQL SnippetC
Validate a pasted read-only SQL snippet for policy, dialect, pagination, plan risk, and visible table or column metadata.
Inputs: sql (required), params, schema, maxRows, +2 more.
ob_verify_permissionsVerify Database PermissionsB
Inspect current database grants and report whether the account appears read-only.
Inputs: profile.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 85 tools
Massive overlap across tool clusters: several tools appear to do nearly the same thing (e.g., ob_explain vs ob_analyze_query_plan, ob_get_database_dictionary vs ob_generate_data_dictionary, ob_generate_er_diagram vs ob_generate_er_document vs ob_generate_er_html). Search and metadata-discovery tools are especially hard to distinguish (ob_search_catalog, ob_search_everything, ob_find_columns, ob_find_tables_by_business_term). An agent would frequently choose the wrong one or need exploration to disambiguate.
All tools share the ob_ prefix and nearly all follow lowercase snake_case verb_noun naming (ob_list_tables, ob_generate_er_diagram, ob_compare_schemas). A few anomalies like ob_policy_explain and ob_config_diagnostics invert the pattern, but overall the convention is predictable.
85 tools is an extreme count for a single database MCP server. Even acknowledging a broad read-only scope, dozens of tools are near-duplicates or micro-variants, making the surface bloated and hard to navigate.
The read-only OceanBase domain is covered exhaustively: connection/health, permissions, querying, pagination, metadata discovery, search, SQL analysis, profiling, data quality, documentation, ER diagrams, and metadata diff/impact analysis. There are no obvious dead ends or missing lifecycle stages for the stated read-only purpose.
Maintenance
Related MCP Connectors
Governed MCP gateway: one endpoint for your tools, with credential custody and audit log.
Read-only MCP access to sessions, funnels, campaigns, errors, live visitors, and anomalies.
Read-only MCP for identity resolution and write guardrails.
Managed MCP gateway with OAuth, per-agent permissions, shared accounts and activity logs.
Related MCP Servers
- AlicenseAqualityCmaintenanceRead-only MySQL/MariaDB MCP server for running SELECT queries safely, with automatic read-only enforcement and query limits.314 npmMIT
- AlicenseAqualityBmaintenanceA read-only MCP server for exploratory data analysis across PostgreSQL, MySQL, and ClickHouse databases, providing safe, read-only access with comprehensive analysis capabilities.1084 PyPI6MIT
- AlicenseNot gradedqualityDmaintenanceA secure, read-only MySQL database proxy using MCP protocol, enabling SQL queries and table inspections via HTTP.14 npm1MIT
- FlicenseNot gradedqualityDmaintenanceA read-only MCP server for Oracle databases that enables SQL queries, schema inspection, and data sampling without requiring OCI client libraries. It supports both TNS alias and direct connection modes with robust security guardrails.-