Atlas / MCP servers / ratel-ai / Ratel

RatelCAUTION

Name: Ratel
Author: ratel-ai

mcp/ratel-ai/ratel

MCP server by ratel-ai · accuracy · agents · claude skills · Audited 2026-10-01

Context engineering for AI agents. ~80% fewer tokens. Fix tool overload. Skills and memory with in-process BM25 and semantic retrieval. Progressive Disclosure. No vector DB.

Verdict

CAUTION

Grade

Trust score

86 /100

Exposed tools

75 63r · 12w · 0d

Transport

streamable-http

License

MIT

Stars

462

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

Ratel Your AI agent is paying for tools it never uses. Ratel fixes that.

Docs • Skills • Discord

Introduction

The context engineering layer for AI agents. Selects only the tools and skills relevant to each turn, recovering accuracy lost to tool overload and cutting what you pay per call. No vector DB, no infra.

Why

Cost: Every tool schema, every skill, and a growing list of instructions in the system prompt are tokens you pay for on every call. Send them all up front and you pay for them all, every turn.
Accuracy: Models get worse as that context grows. Crowd it with tools, skills, and instructions a turn doesn't need and the model picks the wrong option and drifts off task.
Ratel fixes both: it indexes your tools

Read from source at commit 872fff5b5ec2OBSERVED · 2026-10-01

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.

claude-code

claude mcp add telemetry --env ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY} --env AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID} --env AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY} --env AWS_SESSION_TOKEN=${AWS_SESSION_TOKEN} -- npx -y @ratel-ai/[email protected]

claude-desktop

{
  "mcpServers": {
    "telemetry": {
      "command": "npx",
      "args": [
        "-y",
        "@ratel-ai/[email protected]"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
        "AWS_ACCESS_KEY_ID": "${AWS_ACCESS_KEY_ID}",
        "AWS_SECRET_ACCESS_KEY": "${AWS_SECRET_ACCESS_KEY}",
        "AWS_SESSION_TOKEN": "${AWS_SESSION_TOKEN}"
      }
    }
  }
}

Exposed tools (75)

63 read · 12 write · 0 destructive.

Tool	Risk	Description
`a`	read	always one
`address`	read	Shop location
`after_empty`	read	tool listed after an empty page
`after_empty_cursor`	read	tool listed after empty-string cursor
`again`	read	another page
`alpha`	read	first page
`api-design`	read	REST API design patterns: resource naming, status codes, pagination.
`auth`	write	Set up login and sessions.
`b`	read	always two
`bad`	read	an invalid fact
`beta`	read	first page
`boom`	read	always fails
`broken`	write	A native-shaped tool missing its execute handler.
`broken_watch`	read	Watch a broken job.
`cancellation-policy`	read	How to cancel or reschedule a booking and get a refund.
`cap`	read	a capability tool
`charges`	read	...
`ci-triage`	read	Diagnose why the build failed in CI
`deploy`	write	Deploy an app
`deploy-fact`	write	Original fact
`deploy-skill`	write	Deploy an application.
`deploy_app`	write	Deploy the app to production servers.
`docker_build`	read	Build a Docker image from a Dockerfile
`dup`	read	first listing
`echo`	read	echoes its args
`endless`	read	never finishes
`fact`	read	composedonlyterm
`fail`	read	Fail a job.
`first_page`	read	page one
`frontend-slides`	read	Build animation-rich HTML presentations from scratch.
`fs`	read	filesystem helpers
`gamma`	read	second page
`gh_run_list`	write	List CI workflow runs and whether the build passed
`github`	read	GitHub API
`good`	read	a valid fact
`hours`	read	Opening hours
`invalid-number`	read	Invalid number
`lint`	read	Lint the code
`migrations`	read	Reversible DB migrations.
`min`	read	a minimal skill, no tags or body
`mixed_case`	read	Mixed-case schema keys.
`mutable`	read	Accepted description
`n`	read	d
`orphan`	read	A plain native tool with a JSON schema but no id.
`outer`	read	invokes a nested search
`oversized`	read	Oversized schema
`oversized_tool`	read	Oversized schema
`page_one`	read	first page tool
`page_two`	read	second page tool
`pdf`	read	fill pdf forms
`ping`	read	Ping
`push`	write	Deploy the project to Vercel production.
`r`	read	retrieved one
`read`	read	Read a file
`read_file`	read	Read a file from disk
`replaceable`	read	New numeric tool.
`review_code`	read	Review source
`s1`	read	a playbook
`search`	read	Search the tool catalog for matching tools.
`search_files`	read	Grep across files in a directory using a regular expression.
`send_email`	write	Send an email via SMTP.
`shop-address`	read	Where the barbershop is located and its opening hours.
`shop_address`	read	Where the shop is
`skill`	read	composedonlyterm
`stream_outer`	read	streams after a nested search
`stuck`	read	stuck in pagination
`t`	read	a tool
`t1`	read	read a file from disk
`tool`	read	composedonlyterm
`vault_rotate`	read	Rotate a signing key in the vault
`vercel-deploy`	write	How to deploy to Vercel: env vars, preview vs production, rollbacks.
`watch`	read	Watch a job.
`write`	write	Write a file
`write_file`	write	Write a file
`x`	read	d

Trust audit

CAUTIONgrade B · trust 86/100 Install with care. The audit found things worth knowing before you trust its output.

Layer	What it checks	Result
L0	Provenance & inventory	PASS
L1	Static analysis of the code	WARN
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	PASS
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (2 observation(s))
Network: declared (9 observation(s))
Shell: declared (1 observation(s))
Dependencies: not all pinned
Secrets in source: none-found

Findings (20)

MEDIUMCode injection · code.dynamic_import · CWE-78, CWE-94, CWE-95

src/sdk/python/ratel_ai/mcp.py:70

types_mod = __import__(module_name, fromlist=["PaginatedRequestParams"])

MEDIUMNetwork egress · net.beacon_words · CWE-200, CWE-319

src/core/src/usage.rs:4007

"beacon", "harvest", "prune",

LOWInventory / provenance · inv.hidden_file · CWE-1104

.adr-dir

.adr-dir

Why it matters. hidden member outside the usual dotfiles

Fix. review its purpose

LOWInventory / provenance · inv.hidden_file · CWE-1104

.lycheeignore

.lycheeignore

Why it matters. hidden member outside the usual dotfiles

Fix. review its purpose

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

.github/workflows/ci.yml:408

mcp opentelemetry-sdk -e ../../telemetry/python

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

.github/workflows/ci.yml:443

'opentelemetry-sdk>=1.41,<1.42' ../../telemetry/python dist/*.whl

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

src/adapters/ts-mastra/src/package-layout.test.ts:20

new URL("../../../../.github/workflows/release.yml", import.meta.url),

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

src/core/src/usage.rs:4345

"../../../protocol/v1/conformance/vectors.json"

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

src/core/tests/trace.rs:12

serde_json::from_str(include_str!("../../telemetry/conformance/fixtures.json")).unwrap()

LOWNetwork egress · net.raw_ip · CWE-200, CWE-319

src/adapters/ts-vercel-ai-sdk/src/coexistence.test.ts:138

baseUrl: "http://127.0.0.1:1",

LOWNetwork egress · net.raw_ip · CWE-200, CWE-319

src/sdk/python/tests/conftest.py:66

yield f"http://127.0.0.1:{port}/embeddings"

LOWNetwork egress · net.raw_ip · CWE-200, CWE-319

src/sdk/python/tests/conftest.py:105

f"http://127.0.0.1:{server.server_port}/embeddings",

LOWNetwork egress · net.raw_ip · CWE-200, CWE-319

src/sdk/python/tests/test_embedding_artifact.py:135

yield f"http://127.0.0.1:{server.server_port}/embeddings", state

LOWNetwork egress · net.raw_ip · CWE-200, CWE-319

src/sdk/python/tests/test_intent_graph_storage.py:690

return f"http://127.0.0.1:{listener.getsockname()[1]}", listener

LOWSupply chain · supply.unpinned · CWE-829, CWE-1357

examples/adaptive-ranking-ts/package.json

@types/node, tsx, typescript

Why it matters. 3 dependency range(s) float