Use cases
Ten jobs, each written as the chain it actually is: what goes in, which app takes the work next, what comes out. Every app named here ships in the current release and has its own page in this reference.
The short version of this page lives on the site at ownyourmodels.xyz/cases. This one links each step to the documentation for the app that performs it.
1. A 200-slide course built from the standard it teaches
Section titled “1. A 200-slide course built from the standard it teaches”Who: course designer.
Today. The script lives in one tool, the slides in another, the narration in a third. Correct one fact in module two and the export gets rebuilt three times, by hand.
With M.A.V.E.N. One app carries the course from the source document to the narrated deck. Fix the fact once and rebuild. Timings are measured from the audio, not estimated. The SCORM package carries its own player and reports completion to any LMS.
Chain: D-Forge converts the standard → MD Generator repairs the markup → P-Forge outlines the modules → slides and visuals → narration → SCORM package for the LMS.
2. The archive nobody reads becomes something you can question
Section titled “2. The archive nobody reads becomes something you can question”Who: operations lead.
Today. Regulations, manuals and inspection reports pile up as PDFs. Finding one clause means Ctrl+F across a hundred files, and half of them are scans with no text layer.
With M.A.V.E.N. Documents are read, the mess left by OCR is repaired, and the result is indexed. Answers come back with the page they were taken from, so a claim can be checked instead of trusted. Nothing leaves the machine the files already sit on.
Chain: MinerU reads the scans → MD Generator repairs the markup → N-Forge indexes the sources → ask with citations → the answer lands in the vault.
3. Telegram bots that answer for you, on your own keys
Section titled “3. Telegram bots that answer for you, on your own keys”Who: small business owner.
Today. Either a developer is hired for every bot, or a builder is rented where the logic belongs to somebody else and the bot token sits on their server.
With M.A.V.E.N. Each bot gets a profile, a model and a gateway service of its own. Tokens are encrypted on your disk. A pipeline can take a photo sent to a bot and send back a 3D file.
Chain: Hermes Swarm connects the bot → the encrypted token store keeps the token → the profile picks the model → the node graph wires the pipeline → the gateway runs it as a service.
4. Five clients in one workspace instead of “100500” subscriptions
Section titled “4. Five clients in one workspace instead of “100500” subscriptions”Who: marketer.
Today. One subscription for competitor research, another for creative, another for planning, another for reporting. Each bills per seat and calls the same models you are already paying for.
With M.A.V.E.N. Research, angles, creative and the SEO brief run in the same place, one project folder per client, on the machine that holds the client data.
Chain: Ad-Forge runs Ad Scout recon over the Meta Ad Library → R-Forge researches the market → Ad-Forge writes angles and copy → M-Forge makes creative and thumbnails → SEO-Forge produces the brief and reads Search Console.
5. From a paragraph to a shot list, props and a cut
Section titled “5. From a paragraph to a shot list, props and a cut”Who: indie studio.
Today. Storyboards are drawn by hand or skipped, 3D goes out to a contractor, and the edit happens somewhere else again. The look drifts between the three.
With M.A.V.E.N. The script, the boards, the shots, the props and the cut sit next to each other, so a change to the script is one pass, not three briefs.
Chain: Chat writes the script → S-Forge storyboards the scenes → Veo generates the shots → G-Forge models the props → M-Forge cuts and scores.
6. Twenty agent terminals, one board and one bill
Section titled “6. Twenty agent terminals, one board and one bill”Who: developer running agents.
Today. Agents work in tabs. Nobody remembers which one is on what, a closed laptop kills the run, and the cost arrives at the end of the month as a surprise.
With M.A.V.E.N. Work goes on a board, every run is a background job that survives a reload, and spending is visible per engine while it happens. A daily ceiling stops a runaway before the invoice does.
Chain: Kanban holds the task → an agent console (Claude, Codex, Grok) takes it → the job runs in the background → Telemetry shows the cost per engine → the spend ceiling stops the day.
7. A market read you can defend two weeks later
Section titled “7. A market read you can defend two weeks later”Who: analyst.
Today. Forty tabs, copy-paste into a chat window, and by evening nobody can say where the conclusion came from.
With M.A.V.E.N. Sources are collected, claims are checked separately, what survives is indexed, and several models argue over it before an answer is written down with its sources attached.
Chain: R-Forge searches the web → R-Forge Verify checks the claims → N-Forge indexes what survived → AI Mastermind puts it to a panel of models → the report lands in the vault.
8. One conversation, whichever model answers it
Section titled “8. One conversation, whichever model answers it”Who: anyone who writes to models all day.
Today. Claude in one tab, ChatGPT in another, Gemini in a third. The history is split three ways and lives in three accounts.
With M.A.V.E.N. One chat speaks to every engine this install can reach, the model can change mid-conversation, and the history is a file on your disk.
Chain: Chat takes the question → the engine row routes it to any of nine engines → the Model Manager offers only what is installed → Memory keeps what you asked it to → the day note is saved to the vault.
9. What everyone did, in a log that can be checked
Section titled “9. What everyone did, in a log that can be checked”Who: a team of three to ten.
Today. Everyone has their own AI and their own prompts. Decisions happen in private windows and reach the rest of the team as hearsay.
With M.A.V.E.N Teams. Messages, patches, CI runs, approvals and agent turns are events on one signed log with a hash chain, so the history can be verified rather than believed.
Chain: Teams is used as usual → every event is signed → the hash chain makes edits detectable → the shared timeline is read by everyone.
10. Reach your own machine without opening it to the internet
Section titled “10. Reach your own machine without opening it to the internet”Who: anyone away from their desk.
Today. The choice is usually no remote access at all, or a port forwarded to the open internet and some optimism.
With M.A.V.E.N. Loopback stays as it was. Any other host has to pass a PIN, gets a signed session, and whatever a model generates opens in a sandbox that cannot reach back into the app.
Chain: loopback stays open → the PIN gate challenges every other host → the session cookie is signed → generated pages are sandboxed → the spend ceiling caps the day.
Who this is not for
Section titled “Who this is not for”- You want a chat window and nothing else. A browser tab does that, and that is fine.
- You do not want software running on your own machine. That is what M.A.V.E.N is.
- You need a vendor with an SLA and a support desk. Here the administrator is you.
- You need single sign-on and access reviews. Teams keeps a signed log of what happened, but it does not manage who may do what.