Skip to main content
Glama

Observe which hosts a lane really calls

observe_endpoints

Trace outbound connections from a lane's probe prompt via CONNECT proxy, OS socket tables, and Linux strace for TLS SNI/DNS without decryption. Set run:true to execute; false returns commands.

Instructions

Run a lane's one-line probe prompt (fleet.json observe.argv) and record every outbound connection: a logging CONNECT proxy (every OS), the OS socket table for the process tree (ss / lsof / netstat), and on Linux strace for TLS SNI and DNS names. Nothing is decrypted. Settles the CONTROL plane only - observed hosts end at the TLS edge, never the inference origin. Spends one model call per lane, so it only runs with run:true; without it, it returns the commands it would run.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "lane": "<lane>"
    },
    "name": "observe_endpoints"
  }
}

Input Schema

NameRequiredDescriptionDefault
runNotrue to execute (spends one model call per lane). Default false: plan only.
laneYesLane id from fleet.json (e.g. "W7"), or "all".
locateNoAlso run locate_model on each observed host (default true).
timeout_sNoPer-lane timeout (default 120).

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv1.1.0

TDQS

A4/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the full burden and does well: it discloses that nothing is decrypted, that each lane costs one model call, that run:true is required for execution, and that plan mode returns the commands instead of results. It omits operational prerequisites that matter for this mechanism, such as the elevated privileges strace, lsof, and socket-table inspection typically require.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Three dense sentences, front-loaded with the action and mechanism before the scope caveat and the cost gate. Almost every clause carries information (decryption stance, per-lane cost, dry-run behavior) but the enumeration of three collection mechanisms in one sentence is heavy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a multi-mechanism tool with no annotations and no output schema, the description covers cost, safety (no decryption), plan mode, and scope limits, which is substantial. It does not indicate what the recorded output looks like (per-host list, ports, resolution of SNI/DNS names) nor the privilege requirements, leaving real gaps for an agent deciding whether it can run this successfully.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so each parameter is already documented, and the description largely restates that (run:true spends a model call, plan-only default). The 'locate' parameter's meaning is only indirectly supported by the mention of settled hosts ending at the TLS edge; no new syntax or behavioral detail is added per parameter beyond the schema baseline.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource: run a lane's probe prompt and record every outbound connection, enumerated by mechanism (CONNECT proxy, OS socket table, strace on Linux). It also draws a scope boundary ('Settles the CONTROL plane only - observed hosts end at the TLS edge, never the inference origin') that separates it from data-plane siblings such as trace_endpoint or find_regional_endpoints.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly states the execution gate: 'it only runs with run:true; without it, it returns the commands it would run', which tells the agent when it gets a dry-run versus live observation. It also scopes the tool to control-plane host discovery rather than inference-origin tracing. It does not name a specific alternative tool to use for the inference origin, so it stops short of full routing guidance.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.