Multi-language observability SDK with AI-native telemetry, agent instrumentation, and MCP tool integrations for Python, Node.js, Go, and Java services
AgentTel SDK is primarily a telemetry/instrumentation framework, NOT an MCP server. The repository contains multiple language SDKs (Python, Node.js, Go, Java) for observability enrichment, plus an `agenttel-instrument` Python package marked as 'MCP server for AI-driven instrumentation' in its pyproject.toml. However, the source code provided shows NO actual MCP tool definitions, schemas, or protocol implementations. The 13 tools listed (get_service_health, get_incident_context, etc.) are inferred from a Java file (ToolPermissionRegistry.java) that appears to be a permission/identity system, NOT MCP tool registration. We cannot see: (1) explicit tool definitions with input/output schemas, (2) descriptions in any MCP context, (3) parameter type declarations, (4) error handling patterns designed for LLM consumption. The provided code snippets show package.json, pyproject.toml, and go.mod files, dependency and build configuration, with zero MCP protocol implementation visible. Without access to the actual MCP tool definitions, all per-tool scores must be capped at 0 or 50 (inferred only).
Add an entry to an existing session
Create a new session for diagnostics
Execute a remediation action
Retrieve error analysis data
Retrieve executive summary report
Retrieve incident context information
Retrieve playbook instructions
NO TOOL DEFINITIONS FOUND: Tool list is inferred from ToolPermissionRegistry.java (identity/permissions system), not from actual MCP tool registration. Per hard scoring rule, inferred tools cap at 50; combined with missing schemas/descriptions, all tools score 23-25.
ZERO INPUT SCHEMAS: No JSON Schema objects visible for any tool parameter. Per hard scoring rule: 'If a tool has NO input schema at all: its schema score MUST be 0.' All 13 tools score 0 on schema dimension.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 25 | 2026-07-28+ | v2 |
Retrieve service health status
Retrieve session information
Retrieve SLO (Service Level Objective) report
Retrieve trend analysis data
List available remediation actions
Verify the effect of a remediation action
NO DESCRIPTIONS: Tool descriptions are 1-3 short phrases (e.g. 'Retrieve service health status'). Per hard scoring rule: 'If descriptions are under 20 characters: description score CANNOT exceed 40.' These are 20-40 chars, generic, and provide zero LLM guidance on WHEN to use, WHAT the tool returns, or dependencies. All 13 tools score 0 on description dimension.
NO PARAMETER DOCUMENTATION: No visible tool input definitions show parameter names, types, or descriptions. Even tools with multiple logical parameters (e.g. create_session, execute_remediation) lack formal schema definitions and LLM-facing descriptions.
NO OUTPUT SCHEMA DOCUMENTATION: No return type, field descriptions, or structure definitions visible. LLMs cannot know what fields to expect, how to chain tools, or what data is available for downstream decisions.
NO ERROR HANDLING PATTERNS: No evidence of error classification (retryable vs fatal), recovery guidance, or actionable error messages. Agents will receive bare API errors with no next-step guidance.
MISSING TOOL COMPOSITION DOCUMENTATION: Multi-step workflows (e.g. create_session → add_session_entry → execute_remediation) lack clear sequencing, required fields for chaining, and guidance on dependencies. Agents will struggle to compose workflows correctly.
WRITE/DESTRUCTIVE OPERATIONS LACK CONFIRMATION: create_session, add_session_entry, and execute_remediation are marked WRITE/REVERSIBLE but have no visible confirmation, dry-run, or idempotency patterns. Agents may destructively execute without safeguards.
NO MCP PROTOCOL IMPLEMENTATION FOUND: Provided source code is build configuration (package.json, pyproject.toml, go.mod) and package metadata. Zero MCP server code visible. Cannot verify tool registration, message handling, or protocol compliance.