Use Cases

How teams run on Tandem

Three concrete playbooks — the agency juggling eight clients, the legal firm running AI on privileged data, and the solo operator running a five-channel media company out of one dashboard.

AI Consulting Agency · 5–20 people · multi-client

Eight clients. One window. No context-switching tax.

A boutique AI consulting firm manages eight simultaneous engagements. Before Tandem, each client's context lived across Linear, Notion, Google Docs, Slack threads, tmux sessions, and private GitHub repos. Context switching ate an estimated 40% of billable hours.

The problem

When a client pings at 2pm, the consultant needs the right Claude session, the right codebase branch, last week's decisions, their credentials, and the current deliverable — fast. Finding all five across six tools takes ten to fifteen minutes. Agents, meanwhile, spin up fresh each session with no memory of prior engagements.

How they run on Tandem

  • Each client is a Project. Isolated workspace, kanban, skills, agent config, and vault-scoped wiki.
  • Agents arrive caught up. Prior sessions write to the wiki; the next agent reads it and knows what was decided in March without being told.
  • The Activity page is the billable log. Session durations, task updates, and agent actions are timestamped, exportable, and client-auditable.
  • Browser agent pre-loads client portals with the right credentials scoped to the right project — no cross-contamination.
  • One keystroke switches context. Project picker → code, terminal, tasks, wiki, browser all repopulate together.

Stack consolidation

Before
Linear Notion Google Docs Slack tmux ChatGPT Teams Cursor Toggl
After
Tandem Claude API

A client asks "what did you decide on the auth schema in March?" — the wiki the agent wrote itself surfaces the answer in thirty seconds, with the commit, the rationale, and the links to the Slack thread that no longer exists.

Regulated Enterprise · legal · healthcare · finance

Running AI on privileged data — with zero cloud hops.

A mid-sized law firm wants its associates using AI to summarize discoveries, draft memos, and search case history. Every major AI SaaS is off the table: privilege-breach risk, bar-association flags, and GC sign-off that never comes. Local LLMs help but lack an orchestration layer.

The problem

Competitors using AI are billing discovery summarization thirty percent faster. IT and compliance block ChatGPT Enterprise, Cursor, Notion AI, and every tool that sends prompts to a third-party host. The firm is losing against competitors who either accept the risk or are based in jurisdictions without bar rules on this yet.

How they run on Tandem

  • Installed on each attorney's workstation via signed .deb or Tauri bundle — no browser, no SaaS dashboard to audit.
  • All state is local. SQLite, markdown, browser agent — all run on-device. The only outbound traffic is direct Claude API calls from the attorney's own key.
  • Tandem Security provisions the workstation as a hardened Qubes VM via Tutela — a security posture the bar will accept.
  • The wiki becomes firm memory. Precedents, motion templates, client-specific style — all written by the agent, all staying on disk.
  • Sandboxed agents for privileged-doc review. Every agent action appears in an immutable task timeline that compliance can audit.
  • Zero SaaS vendors sit between the attorney and the model. Claude is the only third party, and only for the calls the attorney explicitly initiates.

Compliance posture

Blocked by IT
ChatGPT Enterprise Cursor Notion AI Gemini Workspace Copilot Business
Approved
Tandem (localhost) Tandem Tutela provisioning Direct Claude API under attorney key

The difference is not "we trust the vendor less." The difference is there is no vendor. The work happens on the attorney's own machine, and compliance can see every action an agent took without asking anyone's permission.

Solo Operator · one-person media company

One person, five channels, one dashboard.

An independent technical writer runs a weekly newsletter, a YouTube channel, an X account, a LinkedIn presence, and a consulting side-practice. The prior stack spanned Notion, ConvertKit, Descript, Buffer, Raindrop, Claude desktop, and tmux — and still required hours of manual copy-paste between tools every publishing day.

The problem

Publishing a single Tuesday article took three hours of tool-switching — research lived in Raindrop, outlining in Notion, draft in a word processor, social copy in a Google Doc, video assets in Descript. The stack cost over two hundred dollars a month. Worse, research done for a piece in February was impossible to find by April.

How they run on Tandem

  • Research captured to the vault via the brain-dump page — links, quotes, half-formed ideas, screenshots — auto-classified by agent into the right wiki page.
  • Editorial calendar as a kanban — idea → research → outlining → drafting → published. Every card links to its source notes in the wiki.
  • Drafts are written in Tandem's file editor with an agent helper that pulls voice and recurring phrasings from prior published pieces in the wiki.
  • On publish, the socials module cross-posts a summary to X, LinkedIn, and schedules a Thread — no copy-paste, no tab-switching.
  • YouTube scripts use the same research pipeline. Audio transcribes via local Whisper; the agent drafts show-notes and a newsletter summary from the transcript.
  • Consulting clients and content live side-by-side in Projects — the operator's entire work life is one dashboard.

Stack consolidation

Before
Notion ConvertKit Descript Buffer Raindrop Claude desktop tmux Ulysses
After
Tandem Claude API Local Whisper

Research never disappears. Every link, quote, and rough paragraph from an April Tuesday shows up in the wiki in October when the next piece needs it — written by the agent, indexed by the vault, owned by the creator, untouched by a third-party SaaS.

Which one is you?

Find out where your stack sits on the AI readiness curve — and what replacing it looks like.

Score Your Stack — 2 min → See Pricing