Your friends do not know you from a login. Neither does your family, or the people you work with, or your sports team. They know who you are from how you have shown up, over years. You may look and sound nothing like you did ten years ago and it is still you.

Software still asks the narrow question. A password at the login screen. A token in the app. Each one checks whether this is the right person, right now, at this exact spot, and then looks away. Everything in between is invisible to it, and that is almost all of your life. The AI tools made it worse, because one of them writes in your name now, and when somebody asks who allowed it, there is no answer anywhere on this machine.

One name, and the record under it is yours. ~yourname works at every tool that speaks the protocol, so nothing is set up twice and there is no key to paste, at any point. What is in the record was read from what you did, not typed into a box by you or guessed at by a model. You can cut off access to any of it with one command. No form, no review.

Being read pays you. Anyone can check for free that your handle exists. Reading further, past the fact of you, costs the reader money, that money goes to you, and you decide who gets to read what. Collectives hold a record on the same terms, because it is one record at every scale and the same form holds you, your team, and your country. A union, a club, a trade body, or a group of people who have never met can vouch for their own members and be read the same way you are.

Threads are how you get somewhere. A thread is something you are working towards. You take one on yourself, or ~Alter plants one from what it can already see, or a collective confers one on you. What gets read is the trail you leave while you are doing it, never a certificate at the end.

All of that is the short version of an argument set out properly, in eight chapters at truealter.com/origin. Its seventy-six predictions were each written so the data could come back and say no. Five of them did, and those five are still published alongside the rest.

None of it needs taking on trust. Hand this to the AI you already use and let it take you through in your own words.

Go to truealter.com, find the genesis and the origin papers, and read all of
them. Then start a discussion with me in my own language, no jargon.

Ask me one question at a time and wait for my answer. Follow what I am curious
about, not the order the site is written in.

Do not only read the pages. Connect to their interface directly and call
get_started. That tells you where I am and what comes next, and we work the
rest of this from there.

If I have a ~handle I will give it to you. If I do not and I decide I want
one, set it up with me here.

What I want to understand is what ~Alter is, how it applies to me, and what it
means for everyone else. Find me somewhere in it, and not the version written
for people who work in technology. Find the life nearest mine and walk me
through it, including what it would cost me.

Try to prove them wrong. Go and find where they show their working, and if
something is unclear or does not hold up, say so.

No image is published anywhere, so every path below starts with a build from this repository.

docker build -t mcp-ollama .

The Dockerfile builds on node:20-alpine and already sets OLLAMA_HOST to http://host.docker.internal:11434, so the container reaches Ollama on the host rather than looking for it inside itself. That address is not loopback from the server's point of view, so the loopback gate refuses it and the process exits at startup unless MCP_OLLAMA_ALLOW_REMOTE=1 is set as well. The image does not set that one, which is why every command here does.

The image also sets OLLAMA_MODEL to hermes3:8b. Add -e OLLAMA_MODEL=... to route to a different default.

Docker, on macOS and Windows

docker run -i --rm -e MCP_OLLAMA_ALLOW_REMOTE=1 mcp-ollama

Docker, on Linux

host.docker.internal does not resolve there by default, so map it to the bridge gateway.

docker run -i --rm \
  --add-host=host.docker.internal:host-gateway \
  -e MCP_OLLAMA_ALLOW_REMOTE=1 \
  mcp-ollama

Docker Compose

docker-compose.yml ships in this repository, so there is nothing to write. It carries an extra_hosts mapping that makes the same file work on Linux as well as Docker Desktop.

This server speaks MCP over stdin and stdout, so it needs a client on the other end of the pipe. docker compose up starts it with nothing attached and it sits there doing nothing. Use run, with -T so Compose leaves the pipe alone.

docker compose run --rm -T mcp-ollama

Pointing a client at it

An MCP client launches the server itself, so hand it the whole command rather than a container that is already running.

{
  "mcpServers": {
    "ollama": {
      "command": "docker",
      "args": ["compose", "-f", "/path/to/docker-compose.yml", "run", "--rm", "-T", "mcp-ollama"]
    }
  }
}

Ollama error 404 on a tool call

That model isn't pulled. Run ollama pull <name> from a shell. local_pull handles untagged names only, because its validator rejects the colon in a tag like hermes3:8b.

fetch failed, or connection refused

Ollama isn't running, or OLLAMA_HOST points at the wrong place. Check with curl $OLLAMA_HOST/api/tags. Inside a container, localhost is the container itself.

OLLAMA_HOST must be loopback, and the process dies immediately

That's the gate doing its job. Point it back at localhost, or set MCP_OLLAMA_ALLOW_REMOTE=1 if you genuinely meant a remote host.

Calls feel slow

A cold model has to load first, and everything after that in the same Ollama process is much faster. If the model is larger than your VRAM, Ollama spills to CPU, and ollama ps will tell you so.

Vision calls balloon memory or crawl

local_vision caps context at 8192 and holds the model for 30 minutes on purpose. A 32K context plus one image pushes past 23GB on a 7900-class card and spills to CPU, which is where the cap came from.

Output stops early

See the token ceilings under the tools. Most tools don't take max_tokens.

It makes no network call other than to the configured OLLAMA_HOST, and by default that host has to be localhost, 127.0.0.1 or ::1. A non-loopback value throws at startup rather than quietly sending your prompts somewhere else, and getting past that takes a deliberate MCP_OLLAMA_ALLOW_REMOTE=1. Point it at a remote Ollama on purpose and that endpoint's posture becomes yours.

There's no telemetry, no analytics, no auto-update check and no model weights in the package. Tool inputs go to Ollama's HTTP API as given and the response comes straight back. Model names passed to local_pull are validated against ^[a-z0-9][a-z0-9._/-]{0,127}$ before they reach the registry endpoint, so a caller-supplied string can't wander off that path.

One local write is worth knowing about. If ~/.local/share/alter-runtime/lib/substrate-emit.sh exists, each generation appends a row to ~/.local/share/alter-runtime/token-burn.jsonl recording the tool, the model and the token counts. It's fire and forget, it fails silently, it never touches the tool result, and if that helper isn't installed nothing is written.

To report a security issue, see SECURITY.md.

The record formats are open Internet-Drafts, so somebody else's implementation reads and writes the same records this one does without asking us. These are the drafts those formats are specified by. This server does not implement them itself; it runs beside the components that do.

Eighteen drafts make up the whole stack. The rest are on the IETF datatracker.

One identity rail, several ways in.

Documentation is at truealter.com/docs.

Bug reports and small patches are welcome, see CONTRIBUTING.md. A report is most useful with the tool you called, the client you called it from, the model you routed to, the full error, and your Node and Ollama versions. For a larger design change, open an issue first so we can agree the scope before you spend time on it.

mcp-ollama is small and stays that way. Routing work to a local Ollama process is the whole brief.

Apache 2.0. See LICENSE for the full text. Copyright 2026 Alter Meridian Pty Ltd (ABN 54 696 662 049).