Skip to content
Agents tracked: 502 Releases (7d): 762 Agent status: no reported issuesSince last limit reset: Claude 11d · Codex 9d For AI agents: Markdown Updated Nov 22, 2026

walkerOS Flow

by elbwalker

build and test tracking setups with AI.

Cloud and DevOpsOfficialLocal

GitHubWebsite

1. Made by

“"name": "io.walkeros/flow",”

Published by elbwalker, the company that makes walkerOS, from its own GitHub organization; the registry name io.walkeros and the hosted address app.walkeros.io match walkeros.io.

github.com/elbwalker/walkerOS/blob/dedbdf71653b227f3f61e4c615a6aa4530b89dda/packages/mcps/mcp/server.json

Checked against its repository at one commit, on Nov 10, 2026. Tap a verdict to see its source.How we checkedReport an error

What walkerOS Flow does

walkerOS is an open source way to collect site and app events, such as page views and purchases. It sends them to tools like GA4, ad platforms or a data warehouse. This server lets your AI assistant create that setup, check it, try events on it without sending anything, and bundle it for your site. A hosted version also works on flows saved in the walkerOS app.

What people use it for

Set up GA4 tracking by asking

Ask for a web setup that sends events to GA4, and the assistant creates it, checks it and tries a page view, all on your computer with no account.

Try an event before it goes live

Send a test page view through your setup file and see which tools would receive it, without calling any of them.

Map events to a destination's fields

Let the assistant write the rules that turn your events into the shape a tool such as gtag expects, using the package's own examples.

Learn what a destination needs

Ask what settings a destination, for example Snowplow, needs, and get its options, hints and examples before you write anything.

Bundle the setup for your site

Turn the checked setup into one small script file for your site or server, with only the parts it uses.

Deploy a saved flow

With a walkerOS account and the hosted server, ask the assistant to deploy your checkout flow and report when it is live.

walkerOS Flow popularity

351
GitHub stars, elbwalker/walkerOS
412
downloads last week, @walkeros/mcp
0
public agent projects connect it
walkerOS Flow: downloads per weeknpm, @walkeros/mcp 412
walkerOS Flow: GitHub starselbwalker/walkerOS 351

Tools it gives your agent (8)

What it sends and its license

Sends

real events with their data

To the destinations in your flow (for example GA4 or ad platforms); only when flow_push runs. Opt out: use flow_simulate instead.

Sends

package names and versions looked up

To the npm registry and jsDelivr; on package_search, package_get, and flow_validate when it checks package settings. Opt out: pass offline true to flow_validate.

Sends

anonymous usage events (tool name, outcome, duration, client name, OS, Node.js version)

To app.walkeros.io/api/telemetry (elbwalker GmbH); only after you opt in with walkeros telemetry enable. Opt out: off by default; DO_NOT_TRACK=1 or WALKEROS_TELEMETRY_DISABLED=1 force it off.

Sends

everything the hosted tools read and change in your walkerOS projects

To app.walkeros.io; when you use the hosted server. Opt out: use only the local server.

Free and open source under the MIT license, also for paid products. The hosted server needs a walkerOS account.

In a paid app
Yes

MIT license, free of charge for any use.

“Permission is hereby granted, free of charge, to any person obtaining a copy” source

Pass on its files
With conditions

Allowed when the copyright and permission notice are kept.

“The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.” source

client-work
Yes

MIT allows use in client projects; the hosted server acts with the signed-in person's rights.

“MIT licensed - self-host anywhere, no vendor lock-in” source

Is it alive

Latest version

4.5.0 in the registry, updated Sep 2, 2026

Last release

@walkeros/mcp 4.9.0 on npm, published 2026-11-04 (the registry entry still lists 4.5.0)

Remote endpoint
yes

Alternatives

walkerOS Flow This page

351 starsLocal

posthog

Pick it if you want one hosted analytics product with session replay and feature flags

Local

segment-analytics-node

Pick it if your events already go to Segment

Local

Other servers from the same repository

walkerOS Source Browser

All cloud and devops servers

Get started

Tell your AI assistant

“Add the walkerOS flow MCP server (npx @walkeros/mcp@4.9.0), create a web flow with a GA4 destination in ./flow.json, run flow_validate and flow_simulate on a page view, and do not call flow_push without asking me.”

Connect it

claude mcp add --transport http walkeros https://app.walkeros.io/api/mcp

The local server needs no environment variables and no account. The hosted server signs you in with OAuth in the browser; a script or CI job uses an automation token (wos_pat_...) as a Bearer header instead.

Mistakes to avoid

Calling flow_push to test a new setup
Use flow_simulate first; call flow_push only on purpose, and preferably for server flows.

flow_push makes real calls to real destinations, and web destinations need browser globals Node.js lacks.

Passing inline flow text as configPath to flow_bundle
Pass a file path or a web address.

flow_bundle reads the value as a file name and fails.

Writing a config loaded with flow_load straight back to disk
Remove the user_data wrapper from each string value first.

Values people wrote come back wrapped, and the wrapper would become part of the value.

Guessing a package name for package_get
Call package_search first and use the exact name it returns.

package_get needs the exact package name.

Validating the structured output of every tool with the MCP TypeScript SDK client
Read the text block or skip output-schema validation for flow_load, flow_examples, flow_bundle and flow_simulate.

Their advertised output schemas leave out fields the results carry, so a validating client rejects the results.

Questions

Do I need a walkerOS account?

Not for the local server. Creating, checking, simulating and bundling flows runs on your computer with no account; deploying saved flows needs the hosted server and an account.

Does simulating send data to Google or other tools?

No. flow_simulate makes no real calls; only flow_push does.

Does the server send usage data to walkerOS?

Only if you opted in with walkeros telemetry enable. It then sends tool names, outcomes and timings, never file contents or flow settings.

Can a saved flow in the walkerOS app be tested locally?

Yes. The hosted server gives a config link that lasts about ten minutes, and the local tools read the config from it.