Everything about GitMir

GitMir gives AI and people an understanding of how your product actually works. It reads your code into a living model of the product — its features, the rules it enforces, the paths your customers take, the states its records move through and how all of these connect — and keeps that model current with every push. You can use that understanding three ways: as a model you call over an API (GitMir GMM-1), as a model that runs on your own machine (GitMir GMM Local), or as Intelligence that your people and AI agents reach in the GitMir account, in Slack and over MCP (GitMir Intelligence and the products built on it).

AI now writes much of the code, and nobody holds the whole product in their head. An agent sees the files it opened while the rule it breaks sits in another area, so a change that looks clean in the diff can still break a customer path. Teams spend senior time explaining the product to every agent, reconstructing how things connect and reviewing changes line by line. GitMir holds how the product works, keeps it current, and gives every person and agent the part a task needs — so changes are checked against the product before they land.

Engineering teams that work with AI coding agents; CTOs and engineering leaders who need to know what a change does before it ships; founders who start from an idea and need to know what to build and who will buy it; enterprises that need the same understanding across many systems, inside their own boundary; and anyone who needs a model that gets long documents, codebases and data right.

They are separate products — each with its own page, plan and price — that share one account, one workspace, one model of your product and one pool of credits. GitMir Intelligence is the base: the living model of the product. On top of it: GitMir Vibe and GitMir Cloud Compute for building, GitMir Business Review and GitMir MeetAI for checking changes and decisions, GitMir GoBots, Market, ICP and Outreach for going from an idea to customers, and the GitMir models GMM-1 (cloud) and GMM Local (your machine) for work that has to be right.

No. GitMir works beside the AI tools you already use: it gives them an understanding of your product over MCP, and the model and the agent stay yours. If you want a model from GitMir itself, GitMir GMM-1 (in the cloud) and GitMir GMM Local (on your machine) work with the same tools through the standard chat API.

No. Search finds code that looks like the question and leaves the model to work out, on every question, how the business logic connects. GitMir holds those connections — the features, rules, customer paths and status flows of the product and what depends on what — built once and kept current. That is why it can answer what a change reaches, including the parts that share no word with the change.

It is the idea behind GitMir and the way it is measured: give the same AI model better understanding and it does better work. In Benchmark 001 the model and its settings were held constant; only what the model was given changed — and with GitMir it scored 92% against 78% for the strongest alternative, on 8.4K tokens of context per question instead of 27.4K.

No. GitMir publishes its results and everything needed to repeat them — the benchmark repository, commit, model, settings, questions and scoring — and keeps the method that produces the understanding. The results are checkable without the method.

Sign up — no card — and you get 10,000 credits. Connect a repository through GitHub: GitMir shows the estimated build in credits, builds the model and keeps it current. Or start without an account: ask the public models of real products in the browser. For the models, create a GMM-1 key in your account, or subscribe to GMM Local and paste the one install command into a terminal.

Yes. 8 real open-source products — Supabase Studio, Codex, cal.com, Atlas, Warpdrive, DroneOps Command, Pronto and Faktorio — are modelled in public at gitmir.com/models. Open one in the browser, ask it questions and look at its drawing and architecture; your AI agent can ask it too, over MCP, with no key.

It does not change your code on its own: only GitMir Cloud Compute carries out work, and only work a person has accepted. It does not decide for people: contradictions it finds — in a document, a meeting or a change — are settled by a person. It describes committed code: work not yet committed and branches it has not read are named as outside the model. It does not use what you connect to train models. And it does not publish how it works inside.

Connect repositories and GitMir keeps a living model of what the product does. People ask and see it, agents get sufficient context over MCP, and every change shows what it reached.

The areas and features of the product, its screens and API routes, the customer paths people walk step by step, the status flows its records move through, the rules it enforces and the records it keeps — and the connections between them: what changes, watches, grants, informs or depends on what. People see it as a drawing and in answers; agents get the part a task needs.

Connect a repository through GitHub. GitMir first runs a quick scan and shows the estimated build in credits; the first repository you connect builds by itself, and later ones start when you confirm the estimate. When the build finishes, the model is ready to ask, to draw and to serve to your agents.

Every answer names the commit it comes from. An agent on another branch sees how far the model is from its own work, and the model catches up with every push to the branch you connect. Every push to the connected branch is read by default, and only what the push touched is read again.

Yes. A project with two or more repositories is read as one: one drawing of where they meet, one question over all of them and one MCP address for the whole project. Each repository keeps its own model and updates on its own pushes. Reading repositories together uses no credits and is on every plan.

Any number of private repositories, on every plan including Free. What limits processing is your credit balance, not a repository count.

How a feature works and why; what a change will break before it merges; what a commit changed in the product; whether a document or specification matches what the product does; which customer paths, rules and records a task touches; and what is risky in the product. Ask in plain words in your GitMir account or in Slack, or let your agent ask over MCP.

Which areas of the product the change reaches — including areas and records outside the diff that a text search would not find — which steps of which customer paths change, where a person should review, and which checks to run. Your agent gets the same list over MCP before it writes the change.

Yes. Save the document in your project and GitMir says, requirement by requirement, what already works, what contradicts the product — shown as what the document says against what the product says — what is new and what needs a question. Only the new requirements are left to estimate.

Yes. Without being asked it flags what is risky in the product — for example actions with nothing recorded about who may take them, statuses that nothing leads out of, or a service that reaches almost every area — in your account, to your agents over MCP and in the review of each push.

It says so. The model names the parts it could not read, those parts cost nothing, and the next update reads them again. You never get a full-looking answer built on part of the code.

Yes. Invite GitMir to a channel or send it a direct message: it answers with product rules, flows and what a change reaches, scoped to what that person can see in GitMir. It is part of GitMir Intelligence at no additional price, and asking is not metered.

Free to start with 10,000 credits. Engineering is $29 per active developer per month with 30,000 pooled credits each; viewers and AI agents add no seat.

Over MCP, the standard way AI tools connect to outside knowledge. Add GitMir's MCP server to your tool once — your GitMir account gives the exact command or button for each tool — and from then on your agent asks GitMir before it changes code and gets the business logic the task touches.

Claude Code, Cursor, Codex, VS Code and any other client that speaks MCP, including your own agents. The GitMir models — GMM-1 and GMM Local — connect to any tool that lets you set the address of a chat API, such as Cline, Continue, Aider and OpenCode.

The rules, customer paths, features and records the task touches and how they connect — including parts that share no word with the task — and the commit the answer is current to. Not the whole repository: the part of the product the task needs, so the agent starts from what the product does instead of guessing at it.

For the Intelligence of your repositories: https://lab.gitmir.com/mcp, or https://lab.gitmir.com/mcp?project=<project id> for a whole project. For GitMir Vibe: https://vibe.gitmir.com/mcp — your agent signs you in. For a public model: https://gitmir.com/mcp/<model>, open without a key.

Each answer an agent gets from GitMir over MCP uses 1 credit; working out what a task touches, where GitMir picks the parts of the product the task is about, uses 10. Questions from people — in GitMir and in Slack — are not metered on any plan.

No. Claude Code, Codex, Cursor and your internal agents add no seat. Their questions over MCP use credits, and model updates after their commits use credits from the same pool.

Yes. Instead of the whole repository, the model gets the part of your product the task needs, already connected. A small model then works from far less context, which makes each answer cheaper and faster.

Yes. With GitMir Vibe each change is checked against the product before the agent commits it; GitMir Business Review reports what each push changed in the product; and work done by GitMir Cloud Compute comes back with its checks and with how far the change went in the product.

The account's access and usage terms apply, and there is no per-agent charge. Running GitMir only through agents, with no developer seat, is set up with a person: write to us from the Enterprise page and we confirm the terms for your setup.

GitMir’s own model: more correct answers about long specifications, rule sets and documents — faster, at a lower price per task, over the standard chat API.

Work that has to be right on large inputs: long specifications and rule sets, contracts, codebases, tables, JSON and logs — including inputs larger than ordinary models can take at once. It returns more correct answers, faster, at a lower price per task than a general model reading the same input.

Any tool that lets you set the address of a chat API: Cline, Continue, Aider, OpenCode, the openai libraries for Python and JavaScript, LangChain, or plain HTTP. You give it the address https://app.gitmir.com/v1, a key from your GitMir account and the model name gmm-1.

On generated documents, codebases, tables and logs whose correct answers are known in advance. GMM-1 and a general model got the same input and the same task, and an answer counted only when it matched the known answer exactly. Each benchmark names its size and date; Experiment 002 is published in full on the benchmarks page.

GMM-1 answers knowing your product, and tells you before it starts when a request goes against a rule of your product. Name the project by its address, or set project in your client to its name or id — GitMir GMM-1 in your account lists your projects.

1 credit per 1,000 tokens you send and 5 credits per 1,000 tokens of the answer and the model’s work, at least 1 credit a request. A short question costs about 1 credit. A new account starts with 10,000 credits.

Requests and answers are not stored and not used for training. So that follow-up questions are faster, what GMM-1 reads — a document, pasted code or a data table — is kept in memory for up to 45 minutes after its last use and then erased. Files your agents write into a project that has no repository yet are kept only when an Admin turns on “Keep what agents write” for that project: they stay until you delete them, one by one or all at once, and reading them into the project's model uses credits.

A short question: about a second. A first question about a large document: one to three minutes, with progress streamed while it works; follow-up questions about the same document are several times faster. Tables, logs and JSON of more than a million tokens answer in under a minute; a first question about a codebase or a document that large can take up to 15 minutes.

Yes. Paste the files and ask: who calls a function, what a change reaches, or ask for the change itself — you get the new code of the function and of every part that calls it. On generated codebases of 60, 200 and 500 functions GMM-1 was right on 10 of 10 tasks; a general model reading the same input was right on 10, 9 and 8, and took 7 to 14 times longer. GMM-1 works with what you put into the request; it does not build or run the code.

Yes. Paste a CSV, a JSON export or a log and ask for counts, totals, averages, top lists or joins. The answers were exact in our tests at 300 rows and at 20,000 rows, in under a minute.

Yes. Send the document and one task: a table or JSON of everything it holds, a new version with its own corrections applied, a note for a team, a translation, or a contract with its contradictions settled by the contract’s own order of precedence. What you did not ask to change stays word for word, and contradictions that no rule settles are left as they are and listed for you.

GitMir GMM-1 runs in the GitMir cloud: you call it over the API and pay per token in credits. GitMir GMM Local is GitMir's 30B model running on your own machine: your data never leaves it, there is no charge per token, and you pay $29 per machine per month.

GitMir’s 30B model with GitMir reasoning, running on your own hardware: it works through inputs far beyond its context window, and your data never leaves your computer. Any tool that speaks the standard chat API connects to it.

Subscribe in your GitMir account. The GMM Local screen then shows one command with your license already in it: paste it into a terminal on the machine. It installs GMM Local, switches the machine on, downloads the model (about 18 GB, once) step by step in percent, starts it and says when it answers. Then talk to it with gmm chat, or connect your tools.

No. The model and GitMir reasoning run on your machine, and your inputs and answers stay there. GMM Local talks to GitMir only to renew its license: it sends the license key, a fingerprint of the machine, its operating system and the app version — never what you ask. What it learns about a document stays on your machine for 72 hours; gmm forget clears it.

An NVIDIA GPU with 24 GB of memory or more on Linux, or a Mac with Apple Silicon and 32 GB of memory or more; about 20 GB of free disk. It also runs on the processor with 32 GB of memory, slowly. Run gmm doctor to see what your machine can do before anything large is downloaded.

Whole documents, codebases and data — far more than a model takes in at once. It is answered in full, not cut to fit: in our tests a 2,000-requirement specification was answered exactly, all 16 requirements of a chain.

Yes. In gmm chat, drag a file or a folder into the terminal window — or type its path — and say what to do with it. GMM Local reads it on your machine and gives the model its full text: PDF (on a Mac, scanned pages too), Word, RTF, ODT, HTML, Excel, PowerPoint, EPUB, CSV, plain text and code, or every document in a folder. On Linux, reading PDFs uses poppler-utils, which installs with one command.

Yes. Say where in your message — “save to ~/Desktop/summary.docx”, or a folder — or type /save with a path after the answer. It writes .md, .txt, .docx, .rtf, .csv, .json and .html, and .pdf on a Mac; ask for several files and it writes each one into the folder.

Anything that lets you set the address of a chat API: Cline, Continue, Aider, OpenCode, Codex, the openai libraries for Python and JavaScript, or your own code. The address is http://localhost:7331/v1 and the model is gmm-30b; /v1/chat/completions and /v1/responses both work.

Yes. gmm run --lan opens it to your local network, and the other computers use it with the key from gmm key. Only the machine that runs the model counts toward the subscription.

Yes, once it is installed and the model is downloaded. The license renews itself whenever the machine is online; it needs to be online at least once a week.

Run gmm update. It installs the newest GMM Local in front of you, with progress in percent; the downloaded model and the license stay.

$29 per machine per month, cancel any time. There is no charge per token: the model runs on your hardware.

Use GitMir GMM-1 in the cloud: the same family of models over an API, paid per token from your GitMir credits.

Local Mind reads your conversations with ChatGPT and Claude Code and keeps what you decided, still want and only assume — each with where you said it — held against your product. Your agents read it before they change anything.

The ChatGPT data export — Settings → Data controls → Export data, the ZIP that comes by mail — and Claude Code sessions, the folder of your project in ~/.claude/projects. From Claude Code, one command makes every new session arrive by itself.

Only the text of your conversations: what you and the assistant said, and “[edited <path>]” when an agent changed a file. Images, files, audio, tool output and subagent sessions stay on your computer. Keys, tokens and passwords are removed before upload, and again when GitMir receives the text. It is processed in GitMir’s cloud.

Only what you said or agreed to counts as your decision; what an assistant proposed waits for your Confirm. When you later say “no longer” or “instead”, the earlier decision is marked cancelled, with the line that cancelled it. When two things disagree without that, Local Mind asks you which one holds — it never chooses for you.

Each decision in force is checked against the model of the repository you choose in GitMir Intelligence — in the product, the product does otherwise, not in the product yet, or not clear from the model — and again after each new reading. “Since your last visit” says what changed and which of your decisions it touches.

On the map. Your decisions are drawn by area of your product, always in the same place; open an area to see what you want, decided and assume there with the parts of the product they are about, then open a decision to see it as a chain — what it serves, the part, what changed since your last visit and what the product does now. A timeline shows week by week what you decided and dropped and how the product’s answer changed, and Then / now puts your last visit beside today. The next visit opens where you stopped.

Over MCP. gitmir_decisions gives your agent the decisions in force about a part before it changes it, and what you no longer want; gitmir_decide writes down a decision you just told it, as a proposal you confirm. Your own keys see your decisions; teammates and their agents see only what you shared.

Not until you share — one decision, or all in force. A shared decision shows to everyone who reads that project, with your name, and you can stop sharing at any time.

10 lines of conversation analysed are 1 credit, counted over everything you add; a conversation you already added only adds its new lines. An agent’s question is 1 credit, like any MCP question. 10,000 credits to start; credits are shared with your team’s other GitMir products.

Any conversation, or everything, from the Local Mind screen. What came only from a deleted conversation goes with it.

GitMir as one MCP server in Claude Code, Cursor, VS Code or Codex: your agent asks the product before it writes, and each change is checked against it before it is committed.

Add one MCP server to Claude Code, Cursor, VS Code or Codex. Your agent signs you in, sends you to GitHub to pick the repository and starts reading it.

The rules, customer paths and features the task touches and how they connect — including parts that share no word with the task — so it changes the code knowing what the change reaches.

Before your agent commits, it checks the change against the product: what the change reaches and which rules it touches. Each connection shows its mode in GitMir: with the optional commit hook on in a clone, an unchecked commit is stopped (a person can still skip it with git commit --no-verify); without the hook the check is advisory — your agent is asked to run it; on the Free plan, or before GitMir has read the repository, nothing is checked and GitMir says so. The check is part of GitMir Vibe ($9 per person per month) and the GitMir Engineering seat.

Free to start with 10,000 credits, then $9 per person per month; each person adds 3,000 credits a month to the team's credits. A GitMir Engineering seat includes GitMir Vibe.

A Remote Worker for your project: GitMir carries out the work your team accepts — from the same Intelligence — and commits it to your repository. No per-token billing.

Only work a person has accepted on the project's board.

A person. Work is written down on the project's board with what must be true when it is done, and an Admin of the project accepts it — nothing is carried out before a person does. A person also checks what comes back.

In GitMir's own sandbox, with context from the project's Intelligence. The result is committed to your repository, straight to the branch or as a pull request.

A commit with its checks and what it changed in the product. The task shows what was asked, what was done, the checks and the commits — and how far the change actually went in the product.

No. GitMir Cloud Compute is a Remote Worker priced per project, not per token. Only model updates after its commits use credits, like any other push.

$99 per project per month, however many repositories the project holds. Model updates after its commits use credits.

Each push is compared with the product model: which processes, status flows, rules and records changed, what now depends on what, and which problems the push brought or settled. Computed from two versions of the model, so there is no opinion a person cannot check.

A code reviewer reads the diff and writes an opinion about it. Business Review compares two versions of your product model and reports what changed in the product — the rules, status flows, customer paths and records — and which problems the push brought or settled. Nothing in it is written by a model about your commit.

It speaks about pushes, not single commits. Every push GitMir reads gets a line: a review when the product changed, “No product changes detected” when it did not, and the reading’s state while it runs or if it stops. The first reading of a repository has nothing to compare with, so it has no review.

Preview — free for now: the review itself costs nothing. Reading the push uses credits, as every model update does.

People talk; GitMir MeetAI draws what they mean, raises contradictions as they are said and holds the conversation against your product's GitMir Intelligence model. The participants see it with a link.

Start a meeting on the project it is about, then listen in the browser, paste a transcript or, where the notetaker is set up, invite it to your call. The meeting is held against the model GitMir already has of that product — there is nothing to prepare.

A person. MeetAI raises contradictions as they are said; the people in the meeting decide.

Each decision is recorded with who made it and becomes a task on the project's Work board.

The drawing of the meeting, through a read-only link.

$49 per project per month; analysis uses credits.

Describe a product and GoBots reads it into a GitMir Intelligence model, researches the market with sources, writes requirements that say why they exist and turns the approved ones into tasks on the board.

No. GoBots starts from a description of the product and builds its model before any code exists.

They become tasks on the Work board, where a person approves them.

Idea model & intent — free; reading the idea uses credits. The requirements and the build plan are $99 per venture per month.

Describe a product and GitMir Market finds the alternatives buyers use, the pains and gaps by segment and the opportunities — every finding with its source — then proposes narrow positionings.

From web sources. Every finding comes with its source, so you can check it.

From the market research and the positioning you choose, GitMir ICP works out the customer and finds matching companies on public pages, each scored with the reasons for and against.

Companies. Each one is scored with the reasons for and against it.

Messages written from each qualified company's evidence, approved by you and sent from your own account. Replies come back as evidence for the product.

No. You approve every message and send it from your own account.

Replies come back as evidence for the product.

Each is a separate product at $49 per venture per month (Market $49, ICP $49, Outreach $49); research runs, searches and writing messages use credits. GitMir GoBots, at $99 per venture per month, takes an idea through all of them.

Connect your GitMir Intelligence to the ChatGPT you already use: answers start from your product, market research, customers and decisions, and ChatGPT can check its answer against them.

It is coming soon. People on the waiting list at gitmir.com/products/chatgpt get access first.

Pricing will be announced before access opens.

No. When it opens, it will answer from your product’s GitMir Intelligence — you can prepare that now if you want.

Yes. Free costs $0 with no card: 10,000 credits to start and 1,000 more every month, any number of private repositories, the public models, and GitMir in Slack. How much you can process is bounded by your credits.

Credits pay for the work GitMir does for you: building and updating the models of your private repositories, answers to AI agents over MCP (1–10 credits each), GitMir GMM-1 tokens, research runs in GoBots, Market, ICP and Outreach, meeting analysis in MeetAI and reading each push for Business Review. Questions from people, in GitMir and in Slack, are not metered on any plan.

$29 per active developer per month, with 30,000 pooled credits per paid developer each month, private product models and change impact, MCP access for supported tools and agents, GitMir in Slack and GitMir Vibe. Read-only viewers do not add a seat charge.

People who do active work: developers, and product, BA and QA people who create documents and tasks. Read-only colleagues and clients view for free, people who only ask in Slack need no seat, and AI agents never take one.

Yes — credit packs, bought once, on any plan, and the more you buy the less each credit costs: Basic — 30,000 credits for $7; Pro — 150,000 credits for $29; Max — 600,000 credits for $99.

No. Before a repository is modelled, GitMir shows the estimated build in credits, and processing stops when your balance runs out — it does not run past it. There is no time-boxed trial: the free boundary is credits, not days.

Each is a separate product with its own plan: GitMir Vibe $9 per person per month; GitMir Cloud Compute $99 per project per month; GitMir Business Review: preview — free for now; GitMir MeetAI $49 per project per month; GitMir GoBots $99 per venture per month; GitMir Market, ICP and Outreach $49 each per venture per month; GitMir GMM-1 paid per token in credits; GitMir GMM Local $29 per machine per month. The full list with what each includes is at gitmir.com/pricing.

No. One account and one workspace hold every product you add, one model of your product and one pool of credits. A team owner adds each product where it is needed.

Enterprise is scoped to your company: the deployment, integrations, identity and access requirements, support, invoicing and procurement terms agreed in your proposal.

That depends on the deployment you choose. Cloud: you authorize a repository and GitMir reads it directly — the fastest start. Private Source: the GitMir Local Connector runs inside your infrastructure and processes only the sources you permit; your code is not transmitted, and the intelligence derived from it is synchronized to GitMir. Enterprise: a private GitMir deployment with a private MCP endpoint, in your VPC, on-premises or in an isolated environment, where nothing traverses a GitMir-operated service.

No. Nothing you connect is used to train models, and nothing derived from it is used to improve the service for anyone else.

No. Your sources and the intelligence derived from them are not sold, licensed or shared with third parties, and nothing of yours becomes public.

Yes. With Private Source the code stays in your environment and only the derived intelligence is synchronized; with an Enterprise deployment everything — including the MCP endpoint — runs inside your boundary and is owned by you. For models, GitMir GMM Local runs entirely on your own machine.

Access is set per workspace, project and repository, for people and for AI agents, with Read, Write and Admin roles. In Slack each person is answered with the access they already have in GitMir. SSO and RBAC are available on Enterprise deployments.

Requests and answers are not stored and not used for training. So that follow-up questions are faster, what GMM-1 reads — a document, pasted code or a data table — is kept in memory for up to 45 minutes after its last use and then erased. Files your agents write into a project that has no repository yet are kept only when an Admin turns on “Keep what agents write” for that project: they stay until you delete them, one by one or all at once, and reading them into the project's model uses credits.

Only what renews its license: the license key, a fingerprint of the machine, its operating system and the app version — never what you ask or the files you give it. Versions before 0.1.6 also sent the computer's name, which GitMir does not keep; gmm update brings the current version. What it learns about a document stays on your machine for 72 hours; gmm forget clears it.

Yes — Benchmark 001: one repository, 50 questions written as real engineering requests, the same model and settings for every tool. GitMir scored 92% answer quality against 78% for the strongest alternative; its answers lost 8% of the score to errors and omissions against 22%, on 8.4K tokens of context per question instead of 27.4K, at $0.07 instead of $0.19 per question and 4.8 s instead of 9.6 s.

Yes. Everything needed to repeat it is public at gitmir.com/benchmarks: the repository, the commit, the model, the settings, the questions and the scoring. Run it against whatever you use today.

On the projects GitMir has run, timing the same workflows both ways: +38% more changes accepted per quarter by the same team; 45 min → 8 min to check one agent task; 69% → 5% rework after the first version; $0.35 → $0.08 AI cost per completed agent action. These are measurements of those projects, not a promise about yours — the AI cost is from Benchmark 001.

GitMir's rule is that every figure is labeled as verified, illustrative or a target, and a measured figure names what was measured and where. Illustrative examples on the site are marked as such.

GitMir works beside them. The comparison pages for Claude Code, Cursor, GitHub Copilot, Codex, Windsurf, Devin, Cline, Tabnine, JetBrains AI, Amazon Q Developer, Augment Code, Sourcegraph, Unblocked, CodeRabbit, Greptile and Qodo show what GitMir adds to each tool, how to connect the two and where GitMir can do the work instead; the tools measured in Benchmark 001 show that run too. All of them side by side: gitmir.com/compare.

Use the calculator at gitmir.com/ai-cost: enter the number of engineers, the loaded cost per engineer and your AI agent and token spend, and it shows what understanding the system costs you — from numbers your payroll and invoices already have.

Models of real open-source products that anyone can open without an account: ask them questions, see their drawing and read their architecture — the areas, screens, API routes, customer paths, status flows and rules, each number taken from the model. Today: Supabase Studio, Codex, cal.com, Atlas, Warpdrive, DroneOps Command, Pronto and Faktorio.

Yes. Each public model answers over MCP at https://gitmir.com/mcp/<model>, with no key and no account.

Yes. Name a public repository with an explicit open-source license at gitmir.com/models/add. GitMir checks the license and whether the repository can be read in full, and writes to you with the decision. Every published model names its source, license and commit, and its maintainers can ask for it to be taken down.

No. Public models are served as answers and drawings; there is no bulk export.

GitMir's global repository of how software works: GitHub indexes software by source code, GitMir indexes it by meaning — product capabilities, customer paths, interfaces, data flows, product rules and market maps. The public models are its beginning, and GitMir Market, ICP and Outreach work on this kind of data.

Private deployment and a private MCP endpoint — in your VPC, on-premises or isolated — identity and access requirements with SSO and RBAC, custom sources and integrations, and support, invoicing and procurement terms. The same understanding of your systems then serves every team, model and agent in the company.

With one measurable workflow: a pilot inside your boundary, measured before and after, contracted against the outcome. Book a demo or write to us from gitmir.com/enterprise.

Yes. Your models can change and your agents can change; the understanding of your systems stays — inside your infrastructure — and serves whichever model is doing the work.

The engineering organization behind MIR DIGITAL — a team with decades of production work across software, automation, commerce, operations and data systems. The founder and CTO is Vladimir Miroshnichenko, with 25+ years of building production software systems; the founder signs the measurements GitMir publishes.

Whether reasoning has to live inside the AI model. GitMir moves memory, structure and state outside the model, so the model becomes one part of a larger system that can observe, act, verify outcomes and learn. Research directions: robotics, pharmaceutical research, scientific discovery and autonomous R&D — at gitmir.com/science.

Book a demo, or write to us at gitmir.com/contact; for company-wide deployments, start at gitmir.com/enterprise.