There are several ways to connect, depending on where the agent runs and who starts the work. Most people need only one or two. Each one is introduced briefly below, followed by a table comparing them; every way has its own page with more detail.

Sign in to set one up →

What the ways in are for

Each way in below connects an agent. What it then does to a card, from the card being written through building, CI and a second agent's check to the deploy, and where the card waits for you, is one page.

How work flows →

The ways in

The agent guide & HTTP API

Paste one prompt that points the agent at this server's guide. The guide explains the boards and gives a curl line for every call, so any agent that can run a shell command can start straight away.

Use it when you want the quickest start, or the agent can make HTTP requests but you can't or won't install anything.

More about Guide & API →

MCP server

This server is also an MCP server. Add it as a connector and the agent gets tools to read and write boards, cards and wiki pages, with no shell needed.

Use it when you use claude.ai, the Claude desktop or mobile apps, ChatGPT, or any other client that speaks MCP. It's also a tidy option for Claude Code.

More about MCP →

The slipdock CLI

A command-line client that wraps the whole API. It keeps the token for you, reads better in a transcript than raw curl, and does a few things nothing else does.

Use it when the agent works in a terminal on your machine, or you want what only the CLI has: syncing your wiki to a folder, files over 10 MB, imports and exports.

More about CLI → Every CLI command →

Skills

Longer instructions installed where the agent looks for them. They cover judgement more than syntax: what to pick up next, what belongs on the wiki, and how to work a backlog unattended.

Use it when you'll work with an agent here regularly and want it to follow the board's conventions without being told. Use them alongside the CLI or MCP.

More about Skills →

Runners

Reverses the direction: the board sends cards to an agent on a machine of yours when something happens, such as a card arriving in a list. The agent works the card and reports back.

Use it when you want work to start without anyone opening a session, such as a list that agents empty on their own, or a queue several machines share.

More about Runners →

Webhooks

An automation can call a URL of yours when something happens on a board. The request carries the card's details, so your own service can react to it.

Use it when you have your own service or workflow tool (n8n, Zapier, an agent behind an HTTP endpoint) and it only needs to know when something happened.

More about Webhooks →

Side by side

Skills and runners aren't separate connections. Skills teach an agent that uses the CLI or MCP, and a runner's agent reads and writes through one of them too. They are listed here because they are what you choose between.

Guide & API MCP CLI Skills Runners Webhooks
Install something on the agent's machine No curl is enough No add the address Yes an escript (Erlang) Partly optional files Yes a small service No nothing
Works from claude.ai, ChatGPT and the phone apps No needs a shell Yes as a connector No needs a shell Partly ChatGPT zips Partly scheduled Claude, ChatGPT No —
Read boards and cards, and search Yes Yes Yes Yes through the CLI or MCP Partly the card it was sent Partly the card's fields
Change cards: move, comment, complete Yes Yes Yes Yes through the CLI or MCP Yes reports back on its card No one way only
Leave and read a card's handoff (where it was left) Yes Yes get_card shows it first Yes Yes read on claim, written on stopping Yes its summary, when it stops short No
Make a card recurring, skip or stop one Yes PUT /api/cards/:id/recurrence Yes update_card's recurrence Yes slipdock repeat Yes through the CLI or MCP Yes through the CLI or MCP No one way only
Snooze a card until a day, or wake it Yes POST /api/cards/:id/snooze Yes update_card's snoozed_until Yes slipdock snooze, slipdock wake Yes through the CLI or MCP Yes through the CLI or MCP No one way only
Colour a card's cover and each edge Yes color, border_top, border_right, … on PATCH /api/cards/:id Yes update_card's color, border_top, border_right, … Yes slipdock edit --color, --border, --border-top, … Yes through the CLI or MCP Yes through the CLI or MCP No one way only
Release a batched board's merged work now (after a red release too, with a warning), and see where its releases stand Yes POST and GET /api/boards/:id/release Yes release_now (check: true to read) Yes slipdock release, slipdock releases Yes only when asked: a card is done on merge Yes a runner's release job fires it No one way only
Read and change what a person sees of a card (their card views) Yes /api/card-views Yes card_views Yes slipdock card-view Yes through the CLI or MCP, when asked Yes through the CLI or MCP, when asked No one way only
Add a skill source, read a skill, enable it on a board pinned Yes /api/skill-sources, /api/skill-entries, /api/boards/:board/skills Yes skill_registry Yes slipdock registry Yes through the CLI or MCP, when asked Yes through the CLI or MCP, when asked No one way only
Get the skills enabled on a board, each with its pinned text Yes GET /api/boards/:board/skills Yes get_guide's board Yes slipdock skills board Yes through the CLI or MCP, when planning Yes through the CLI or MCP, when planning No one way only
Wait on a person, and answer a wait Yes POST /api/cards/:id/wait Yes update_card's wait, answer_wait Yes slipdock wait, slipdock answer Yes open a wait and stop Yes parks the job; the answer sends it again No one way only
See what is waiting on you, and a board's pulse Yes GET /api/work, /api/boards/:id/pulse Yes whoami, get_board's pulse Yes slipdock work, slipdock pulse Yes through the CLI or MCP Yes through the CLI or MCP No one way only
See what in an epic's tree is waiting on you Yes GET /api/cards/:id, /api/boards/:board: waiting_on_me Yes get_card, get_board: waiting_on_me Yes slipdock card, slipdock board: Waiting on you: Yes through the CLI or MCP Yes through the CLI or MCP No one way only
Dismiss an item, or mute a board, on your Waiting on you Yes POST /api/work/dismissed, POST /api/work/muted Yes update_card's and update_board's waiting_list Yes slipdock work --dismiss, --mute Yes through the CLI or MCP, when asked Yes through the CLI or MCP, when asked No one way only
Turn the overseer on, set its pace, and read its alerts Yes GET/PATCH /api/boards/:id/overseer Yes get_board's overseer, update_board's overseer Yes slipdock overseer Yes through the CLI or MCP Yes a runner with job_overseer runs its passes No one way only
See the top cards of lists from several boards (Multiboard) Yes GET /api/multiboard No not a tool: use the API or the CLI Yes slipdock multiboard Yes through the CLI No one way only No one way only
Save a multiboard view by name, for the home page Yes /api/multiboards No not a tool: use the API or the CLI Yes slipdock save-multiboard, multiboards Yes through the CLI No one way only No one way only
Get your addresses for emailing cards and wiki pages onto a board, or rotate them Yes GET /api/boards/:id/inbound, /api/me/inbound (?folder= for a wiki folder's) No not a tool: the address is for a person's mail client Yes slipdock inbound <board>, inbound me, inbound <board> --folder F Yes through the CLI Yes through the CLI No one way only
Set the paths a person approves before a pull request changing them merges Yes PATCH /api/boards/:id sensitive_paths No not a tool: the board's owner decides Yes slipdock set-board <board> --sensitive-paths LIST Yes through the CLI Yes through the CLI No one way only
Let other addresses mail a board's card address (permitted senders) Yes PATCH /api/boards/:id inbound_senders No not a tool: the board's owner decides who may mail it Yes slipdock set-board <board> --inbound-senders LIST Yes through the CLI Yes through the CLI No one way only
See the email you sent in, and why any bounced Yes GET /api/inbound/log No not a tool yet Yes slipdock inbound log Yes through the CLI Yes through the CLI No one way only
Read a board's flow day by day, and how its runners are doing Yes GET /api/boards/:id/flow, /api/boards/:id/runner_stats Yes get_board's history, runners Yes slipdock flow, slipdock runner-stats Yes through the CLI or MCP Yes through the CLI or MCP No one way only
See what is happening right now: which runners are doing what, the cards in progress Yes GET /api/boards/:id/now Yes get_board's now Yes slipdock now Yes through the CLI or MCP Yes through the CLI or MCP No one way only
Record deploys, and read the delivery (DORA) numbers Yes POST /api/deployments, GET /api/boards/:id/dora Yes get_board's dora Yes slipdock deploy run, slipdock dora Yes through the CLI or MCP Yes through the CLI or MCP No one way only
Pause runner work, and see that it is paused Yes POST /api/boards/:id/pause Yes manage_automation's pause Yes slipdock pause, slipdock resume Yes stop when paused Yes nothing is claimed while paused No one way only
Tidy a board up: see what the gardener found, apply or dismiss Yes GET/POST /api/boards/:board/garden Yes garden, garden_decide Yes slipdock garden Yes through the CLI or MCP Yes through the CLI or MCP No one way only
Read and write wiki pages Yes Yes Yes Yes and when to Partly through the CLI or MCP No
Attach a board, card or page to a chat yourself No the agent fetches it Yes as a resource, from the + menu No No No No
Attach and download files Yes Yes up to 10 MB each Yes up to 250 MB, recordings too Yes through the CLI or MCP No No
List and change a board's automations Yes Yes the board's owner Yes Yes through the CLI or MCP No No
Connect runners Yes No Yes Partly through the CLI Partly from the board No
Mirror your wiki to a local folder No No Yes slipdock sync No No No
The board starts the work No you start it Partly claim_job in a loop Partly claim-job in a loop Partly slipdock-loop Yes pushed from a rule Yes on any rule
Runs with nobody watching Partly if you schedule it Partly if you schedule it Partly if you schedule it Yes slipdock-loop Yes Yes
Can be limited to read-only Yes read-only token Yes approve read-only Yes read-only token Yes follows the token Partly its token takes jobs only No —
Signs in with approve a code, or a token the browser (OAuth), or a token slipdock auth whatever the agent uses a runner token nothing: it calls out

yes · partly · no

If you're not sure

  • Using claude.ai, ChatGPT or a phone? Add the MCP connector.
  • Claude Code, Codex or another terminal agent? Start with the guide, then add the skills and either MCP or the CLI once you use it often.
  • Want cards worked without you there? Connect a runner.
  • Only need to tell another system something happened? Use a webhook.