接入你的智能体
从发布者处安装。不是 PlugMCP 端点。
远程 URL 还是本机命令? 两种安装方式的区别
Model Context Protocol (MCP) server for IBM Cloud Logs, enabling AI assistants to interact with IBM Cloud Logs instances. Includes 8 portable Agent Skills for use with Claude Code, Cursor, Gemini CLI, GitHub Copilot, and 30+ other agents.
This project provides two complementary ways to work with IBM Cloud Logs through AI agents:
Key Features:
Option 1: Homebrew (Recommended)
brew tap tareqmamari/tap
brew install logs-mcp-server
logs-mcp-server --version
Agent Skills are automatically installed to ~/.agents/skills/ after Homebrew install.
Option 2: Build from Source
git clone https://github.com/tareqmamari/cloud-logs-mcp.git
cd cloud-logs-mcp
make deps && make build
https://[instance-id].api.[region].logs.cloud.ibm.comus-south, eu-de, au-syd, etc.Configuration file: ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"ibm-cloud-logs": {
"command": "logs-mcp-server",
"env": {
"LOGS_SERVICE_URL": "https://[your-instance-id].api.[region].logs.cloud.ibm.com",
"LOGS_API_KEY": "your-ibm-cloud-api-key"
}
}
}
}
After setup: Restart Claude Desktop and ask "List my IBM Cloud Logs alerts"
Prerequisites:
Configuration Options:
Option 1: Personal Setup (Environment Variables)
# PowerShell (Windows)
$env:LOGS_SERVICE_URL = "https://[your-instance-id].api.[region].logs.cloud.ibm.com"
$env:LOGS_API_KEY = "your-ibm-cloud-api-key"
# Run the server
logs-mcp-server
Option 2: Enterprise Deployment (Microsoft 365 Admin Center)
For organization-wide deployment, configure through Microsoft 365 Admin Center:
logs-mcp-serverLOGS_SERVICE_URL: Your IBM Cloud Logs endpointLOGS_API_KEY: Service ID API key (recommended for enterprise)After setup:
Note: Microsoft 365 Copilot MCP support is currently in preview. Configuration steps may vary by tenant settings. See CONTRIBUTING.md for alternative setup methods.
See CONTRIBUTING.md for setup instructions for Cline, programmatic usage, and other MCP-compatible clients.
88 tools organized by functionality:
query_logs, submit_background_query, get_background_query_status, get_background_query_data, cancel_background_queryingest_logslist_alerts, get_alert, create_alert, update_alert, delete_alertlist_alert_definitions, get_alert_definition, create_alert_definition, update_alert_definition, delete_alert_definitionsuggest_alert - SRE-grade alert recommendations (see Alert Intelligence below)list_dashboards, get_dashboard, create_dashboard, update_dashboard, delete_dashboardlist_dashboard_folders, get_dashboard_folder, create_dashboard_folder, update_dashboard_folder, delete_dashboard_foldermove_dashboard_to_folder, pin_dashboard, unpin_dashboard, set_default_dashboardlist_policies, get_policy, create_policy, update_policy, delete_policylist_outgoing_webhooks, get_outgoing_webhook, create_outgoing_webhook, update_outgoing_webhook, delete_outgoing_webhooklist_e2m, get_e2m, create_e2m, replace_e2m, delete_e2mlist_rule_groups, get_rule_group, create_rule_group, update_rule_group, delete_rule_grouplist_data_access_rules, get_data_access_rule, create_data_access_rule, update_data_access_rule, delete_data_access_rulelist_enrichments, get_enrichment, create_enrichment, update_enrichment, delete_enrichmentlist_streams, get_stream, create_stream, update_stream, delete_streamlist_views, get_view, create_view, replace_view, delete_viewThe suggest_alert tool provides SRE-grade alert recommendations based on industry best practices from Google SRE, the RED/USE methodologies, and academic research on alerting.
suggest_alert?"Suggest an alert for high error rate on my API service with 99.9% SLO"
"What alerts should I create for my Kafka cluster?"
"Help me set up monitoring for my PostgreSQL database"
RED Method (Rate, Errors, Duration) for services:
web_service, api_gateway, worker, microservice, serverlessUSE Method (Utilization, Saturation, Errors) for resources:
database, cache, message_queue, kubernetes, storage{
"suggestions": [{
"name": "API Error Rate - Fast Burn (Page)",
"severity": "P1",
"methodology": "RED",
"burn_rate_condition": {
"slo_target": 0.999,
"burn_rate": 14.4,
"window_duration": "1h",
"consumption_percent": 2.0
},
"suggested_actions": [
"1. Check error logs for patterns",
"2. Review recent deployments",
"3. Verify dependent service health"
],
"runbook_url": "/runbooks/web_service/error-rate"
}]
}
The alerting engine implements recommendations from:
The binary embeds 8 Agent Skills following the agentskills.io open standard. Skills are portable instruction bundles that work across 30+ AI agents — no runtime, authentication, or network required.
Skills are embedded in the binary. Use the skills subcommand to manage them.
# Install to ~/.agents/skills/ (user-level, available to all projects)
logs-mcp-server skills install
# Install to ./.agents/skills/ (project-level, current project only)
logs-mcp-server skills install --project
# List all available skills
logs-mcp-server skills list
# Remove installed skills
logs-mcp-server skills remove
If you installed via Homebrew, skills are automatically installed to ~/.agents/skills/ on first install.
Skills use a progressive disclosure model to minimize context window usage:
This means skills consume ~2K tokens on-demand compared to the MCP server's ~25K fixed overhead per conversation. See BENCHMARK.md for a detailed comparison.
Skills work with any agent that can read markdown files from ~/.agents/skills/ or ./.agents/skills/:
~/.agents/skills/.agents/skills/~/.agents/skills/For maximum effectiveness, use both together — skills provide the domain knowledge, MCP executes the actions.
"Search logs for errors in the last hour"
"Create an alert when error rate exceeds 100 per minute"
"Suggest alerts for my web service with 99.9% SLO"
"What monitoring should I set up for my Redis cache?"
"List all my dashboards"
"Ingest a test log message for my-app"
"Show me all retention policies"
Plus one of the following:
LOGS_SERVICE_URL - Full service endpoint URL (region and instance ID are auto-extracted), ORLOGS_REGION + LOGS_INSTANCE_ID - Region and instance ID (service URL is constructed)Configure multiple IBM Cloud Logs instances:
{
"mcpServers": {
"logs-production": {
"command": "logs-mcp-server",
"env": {
"LOGS_SERVICE_URL": "https://prod-id.api.us-south.logs.cloud.ibm.com",
"LOGS_API_KEY": "prod-api-key",
"LOGS_REGION": "us-south",
"LOGS_INSTANCE_NAME": "Production US"
}
},
"logs-staging": {
"command": "logs-mcp-server",
"env": {
"LOGS_SERVICE_URL": "https://stage-id.api.us-south.logs.cloud.ibm.com",
"LOGS_API_KEY": "staging-api-key",
"LOGS_REGION": "us-south",
"LOGS_INSTANCE_NAME": "Staging US"
}
}
}
}
MCP Server Best Practices:
LOGS_TLS_VERIFY=true)LOG_LEVEL=debug) for compliance trackingAgent Skills Security:
go:embed — immutable after buildibm-cloud-logs-access-control skill includes security query templates (auth failures, privilege escalation, sensitive data access) and compliance patterns (GDPR, SOC 2, multi-tenant isolation)See BENCHMARK.md for a detailed MCP vs Skills security comparison.
See SECURITY.md for comprehensive security guidance.
make deps # Download dependencies
make build # Build binary
make test # Run tests
make lint # Run linters
├── main.go # Entry point with skills subcommand
├── embed_skills.go # Embeds .agents/skills/ into binary
├── .agents/skills/ # 8 Agent Skills (agentskills.io format)
│ ├── ibm-cloud-logs-query/ # DataPrime query writing & validation
│ ├── ibm-cloud-logs-alerting/ # SRE-grade alerting with burn rate math
│ ├── ibm-cloud-logs-incident-investigation/
│ ├── ibm-cloud-logs-dashboards/
│ ├── ibm-cloud-logs-cost-optimization/
│ ├── ibm-cloud-logs-ingestion/
│ ├── ibm-cloud-logs-access-control/
│ └── ibm-cloud-logs-api-reference/
├── internal/
│ ├── auth/ # IBM Cloud IAM authentication
│ ├── client/ # HTTP client with retry/rate limiting
│ ├── config/ # Configuration management
│ ├── tools/ # MCP tool implementations (88 tools)
│ ├── skills/ # Skill installer
│ ├── server/ # MCP server
│ ├── health/ # Health checks
│ └── metrics/ # Metrics tracking
├── Makefile # Build automation
└── .goreleaser.yaml # Release configuration
See CONTRIBUTING.md for detailed development guide including:
Debug mode:
{
"env": {
"LOG_LEVEL": "debug",
"LOG_FORMAT": "console"
}
}
Copyright (c) 2025. All rights reserved.
Model Context Protocol (MCP) server for IBM Cloud Logs, enabling AI assistants to interact with IBM Cloud Logs instances.
工具
暂无工具列表。远程服务器在首次健康检查后会获得工具列表。
常见问题
在本页复制适用于 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 配置。它使用发布者的安装方式,而不是 PlugMCP 代理。
此记录未列出必需的环境变量。
Cloud Logs 通过 stdio 在本地运行。
上次检查时间:2026-10-28。 Model Context Protocol (MCP) server for IBM Cloud Logs, enabling AI assistants to interact with IBM Cloud Logs instances.