MCP outliners compared: Workflowy, Tana Outliner, Checkvist, and listlis

An MCP connection is not a reason to move your outline. It is a reason to ask a better question: what will the AI be allowed to see and change, where will the connection run, and will the result preserve the context that makes a line useful?

Workflowy, Tana Outliner, Checkvist, and listlis all let an AI client read and work with an outline. They make different choices about where the server lives, how writes are controlled, and whether the agent gets a task list, a structured graph, or the branch around a project. Those choices matter more than the letters MCP.

The short answer

Pick Workflowy when you want a local bridge from the desktop app you already keep open. Pick Tana Outliner when supertags and fields organize your graph and you want a choice between free local MCP and paid hosted access. Pick Checkvist when your outline is primarily a task system with assignees, priorities, and dates. Pick listlis when one plain tree needs to carry the project, its notes, its sources, and its next actions without handing the agent a flat search result.

None of those is the universal winner. They solve different problems.

Tool	Where MCP runs	Access and write model	Best fit
Workflowy	In the signed-in desktop app	A generated connection link lets the client read, search, navigate, and update while the app stays open	A local AI workflow around an existing Workflowy account
Tana Outliner	Free local server at http://localhost:8262/mcp, or paid hosted beta at https://app.tana.inc/mcp	Authorize the connection, then use the same read and mutation tools locally or remotely	A structured graph of nodes, supertags, and fields
Checkvist	Hosted at Checkvist	OAuth can be read-only or read-and-modify; MCP requires Pro	Task lists with tags, dates, priorities, and assignees
listlis	Hosted at mcp.listlis.com	OAuth scopes separate reading, writing, sharing, and API-key management	A project tree where the surrounding branch is part of the answer

Workflowy keeps MCP close to the desktop app

Workflowy’s Desktop MCP is the clearest local option. You sign in to the desktop app, enable Local MCP in Settings, copy the generated Connection Link into a compatible client, and keep the app running while the client uses it. The connection needs no separate API key because the desktop app already knows which Workflowy account is signed in.

That is a good arrangement when the computer is where the work happens. A local client can search notes, summarize a project, add tasks from a chat, or reorganize nodes without a hosted connector sitting between the AI tool and the desktop app. Closing the app or disabling Local MCP stops the connection.

It also sets a boundary. The connection depends on that machine and that running app. If you want an agent to reach the same outline from another device or a hosted environment, a local bridge is the wrong shape. Workflowy offers other AI routes, including its local cache tools, but Desktop MCP is the simple answer for a person already working in the desktop app.

Tana Outliner offers local and hosted MCP

Tana Outliner runs a local MCP server inside its desktop app at http://localhost:8262/mcp. It is enabled by default, available on the Free plan, and works while the desktop app is open. A compatible client authorizes through an approval prompt in the app. Personal API tokens are available as a fallback.

Its Remote MCP beta is hosted at https://app.tana.inc/mcp. It uses HTTP and OAuth, works without the desktop app running, and requires a paid Outliner plan. Current pricing starts at $8 per month for Plus, with Pro at $14 per month.

Both servers expose the same tools. An AI client can list workspaces, search nodes, read nodes and their children, and inspect supertags and schemas. Mutation tools can import Tana Paste, create or apply supertags, set fields, edit node names and descriptions, check tasks, create calendar nodes, and move nodes to trash.

Those mutation tools write directly after connection authorization. The proposal-review workflow at https://home.tana.inc/mcp belongs to the separate, newer Tana product.

Tana Outliner makes sense when supertags and fields already carry meaning in your graph. The MCP tools can read that schema and create content that fits it. Local MCP suits work done on one computer. Remote MCP suits clients and agents that need access while the desktop app is closed.

Checkvist makes the connection a task-management tool

Checkvist hosts its server at https://checkvist.com/mcp. Its OAuth flow lets you choose read-only or read-and-modify access, and its Profile → Tools page can disconnect a connected app. That is a useful, legible permission decision. An API-token connection is also available, but it always has full read-and-modify access.

The tool coverage matches Checkvist’s task focus. An assistant can read and search lists, add and update items, work with tags, notes, due dates, assignees, priorities, completion, and Markdown export. MCP access is a Pro feature, so it belongs in the cost decision alongside the rest of Checkvist’s paid task features.

Choose Checkvist if an action list is the thing you need an AI to understand. It is especially credible for work that depends on assigning people, scheduling tasks, and managing a mature list of commands. It is less compelling when the notes, research, and rough project thinking are the reason the tasks make sense in the first place.

listlis gives the agent the branch, not only the match

listlis hosts MCP at https://mcp.listlis.com/mcp. You add the server to a compatible client, sign in through OAuth, and approve the permissions requested. Read, write, sharing, and API-key management are distinct scopes. Connected apps can be revoked from the account menu, which invalidates the grant instead of leaving an old connection alive.

The useful part is the context returned after the connection. listlis_get_context can return a node’s Markdown body, ancestors, siblings, and a bounded subtree. Search results carry breadcrumbs. listlis_pack gathers the relevant branches for a topic into one trimmed Markdown bundle, and listlis_agenda answers what is overdue, due today, or due in the next seven days. The agent can then create a nested outline, edit a node, or move a whole branch; listlis computes the sibling order rather than asking the client to invent it.

Picture a product launch branch with source notes under the feature work, decisions under a meeting line, and follow-ups beneath the decisions. A search result called “pricing page” is not enough. The parent, nearby work, and child notes explain what the phrase means. That is the material listlis gives an AI before it starts moving things around.

Try it with one real branch. Connect an AI client, ask it to use get_context on the branch, then ask it to pack the topic before proposing the next actions. Review the result in the outline. If the agent can identify the work without flattening the project into detached notes, the connection is helping.

Choose the connection after you choose the outline

The MCP protocol is shared ground. The better choice comes from the work you need the outline to hold.

Workflowy is the practical local bridge. Tana Outliner is the structured graph with free local MCP and a paid hosted beta. Checkvist is the task manager with a clear hosted permission choice. listlis is for one keyboard-first tree where sources, notes, tasks, and the context around them stay together, while a connected client can retrieve only the branch it needs.

For the wider product decision, see the outliner comparison hub and the Workflowy MCP comparison. If you want to test the last option, open a root in listlis, connect one AI client, and give it one live project rather than your archive.