llms.txt File

Machine-readable discovery file for AI agents. Helps Claude, LLMs, and AI applications understand Neatlogs capabilities, APIs, and integration options.

The llms.txt file is a machine-readable discovery document that helps AI agents, LLMs, and autonomous applications understand what Neatlogs can do and how to interact with it — the same way robots.txt tells web crawlers what to index.

When Claude or another AI agent encounters Neatlogs, it can read /llms.txt to immediately understand the platform — tools, authentication, endpoints, and workflows — without crawling the full documentation.

Where it lives

Served from the root of the domain, public, no authentication:

Read it from anywhere

curl https://neatlogs.com/llms.txt

In development the file is neatlogs-app/public/llms.txt.

What's inside

The sections follow the order an agent needs them — what is this, what can I do, how do I connect:

#	Section	Content
1	Overview	What Neatlogs is and key capabilities
2	MCP interface	Endpoint, authentication, and connection methods
3	MCP Tools	All 15 available tools organized by category
4	Span Types	Execution types (LLM, TOOL, AGENT, …)
5	Detections	What issues Neatlogs monitors automatically
6	Documentation	Pointers to guides and references
7	API Info	Base URLs and core routes
8	Use Cases	4 real-world agent workflows
9	Rate Limits	Quotas and performance limits

The MCP tool catalog

All 15 tools, grouped exactly as in the file:

Tool	What it does
`ping`	Health check
`whoami`	Get authenticated project ID

Tool	What it does
`search_traces`	Full-text search over traces
`get_trace_context`	Fetch complete span tree with execution details
`list_detections`	List active detection types
`get_detection_trend`	Trend analysis for detections (hourly/daily)

Tool	What it does
`log_trace`	Push an execution trace from the agent to Neatlogs

Tool	What it does
`triage_list`	List all fixes with status
`triage_get`	Get single fix with root cause analysis
`triage_accept`	Accept draft fix to Kanban board
`triage_dismiss`	Reject fix as false positive
`triage_update_status`	Move fix: draft → open → in_progress → resolved
`triage_update`	Edit fix fields (title, severity, label, action)
`triage_get_investigation`	Get investigation thread for fix
`triage_list_steps`	List AI analysis steps for a fix

Detections

Everything Neatlogs monitors automatically, by group:

Detection	Catches
`hallucination`	False or fabricated content
`prompt_regression`	Quality degradation over time
`silent_degradation`	Gradual quality loss
`tool_contract_misuse`	Tool called with wrong arguments
`retrieval_miss`	RAG failure (no relevant results)
`orchestration_inefficiency`	Unnecessary steps in execution
`guardrail_bypass`	Safety constraint violation

Detection	Catches
`cost_anomaly`	Token usage / spend spike
`latency_anomaly`	Response time degradation
`error_detected`	Exception or error in execution

Detection	Catches
`tool_failure`	External API failure

How an agent uses it

Discover — the agent finds https://neatlogs.com/llms.txt and reads it in one request.

Understand — it learns the capabilities, the MCP endpoint, and the available tools.

Authenticate — it sees that access works with project API keys, and how to send them.

Decide — the use-case section tells it whether Neatlogs is relevant to the task at hand.

Act — it calls MCP tools directly, following links into the docs only when it needs depth.

Without llms.txt, an agent has to crawl documentation, guess at endpoints, and discover authentication by trial and error — slow, token-hungry, and unreliable.

Format

Plain text with Markdown structure — human-readable for developers, machine-parseable for LLMs:

llms.txt

# Neatlogs - AI Agent Observability Platform

## Overview
Neatlogs is an observability platform for AI agents and LLMs.
It captures execution traces, detects issues (hallucinations,
regressions, anomalies), and provides AI-powered fixes.

## Primary Interface: MCP (Model Context Protocol)

### Endpoint
https://neatlogs.com/mcp

### Available MCP Tools (15 tools)
…

When it's useful

AI agent exploring the platform

Gets an immediate overview of capabilities in one request.

Claude using the API

Discovers MCP tools and schemas without manual lookup.

IDE plugin integration

Understands authentication and rate limits before the first call.

Autonomous agent triage

Reads the use cases to decide whether Neatlogs solves its problem.

Keeping it current

The file is only useful while it's true. When something ships, update the matching section:

Change	Update
New MCP tool	Tool list
New detection type	Detections section
Docs restructure	Documentation links
Rate limits adjusted	Quotas section
New agent workflow	Use cases

Do: keep tools and limits current, link to accurate docs, use real use cases, state authentication clearly.

Don't: include secrets, leave broken links, or let tool and endpoint information go stale — agents act on whatever the file says.