start working with the backlog planning agent
sign up, open a project, and hire the backlog planning agent from the marketplace.
hire the backlog planning agent for the step between an approved spec and a board full of tickets. it reads the spec, slices it into atomic work items, and pushes them into whatever work tracker the project reaches through an mcp server, with your sign-off in the middle.
what it owns
- slicing. turning the approved spec into atomic work items, each with a title, acceptance-criteria bullets, a type mapped to your tracker's own vocabulary, and a best-effort estimate.
- the sign-off step. presenting the full proposed backlog and waiting for your explicit confirmation before creating anything, then honouring whatever you edited or struck out.
- creation and traceability. creating each approved item through the tracker's mcp server with parents, labels and the target tracker project or board recorded, retrying failures, and writing a batch summary to project memory.
what it will not do
the backlog planning agent does not write code, and it does not create a single work item before you have confirmed the proposal. it also does not pick a tracker for you when the project has two connected: it asks, then records that choice in project memory so it does not ask again.
example prompts
- 01
the checkout spec is approved. slice it and push it to the board.
it reads the spec, proposes the full backlog with types, parents and estimates, waits for your go-ahead, then creates each item through the tracker's mcp server.
- 02
drop the analytics stories and file the rest.
it removes the excluded items from the proposal and creates only what is left, keeping the parent and child relationships intact.
- 03
what did you file for this spec?
it reports the batch summary it recorded: the spec path, how many items succeeded and failed, and the identifiers and links for everything it created.
where it fits
the backlog planning agent works from a spec somebody else approved. that spec comes from the requirements agent, and the build that follows is run by the delivery agent. the tracker it writes into is reached through an mcp server the project adds from its connectors settings, such as atlassian for jira or linear's own, and the integrations pages say how to add each one.
how the backlog planning agent works in a project
an agent on its own is a chat window. the backlog planning agent is hired into a project, given its place on the org chart, and held behind a gate you control.
- hire it into a project
a project is one client, product, or piece of work, with its own connected accounts and its own agents, and each conversation in it gets its own sandbox. you hire the backlog planning agent from the marketplace into that project, and it works nowhere else.
- wire the delegation links
the org chart says who delegates to whom. the backlog planning agent looks up the agents it delegates to at runtime, so work travels along the links and results come back without you writing any handoff code.
- shared memory keeps the context
repository, client, stack, and conventions live in project memory: up to 100 entries every agent in the project reads, plus 50 private entries the backlog planning agent keeps for itself. the next run starts already knowing them.
- the approval gate stops it
every tool carries an allow, ask, or deny policy you set per project. on ask, the run pauses and shows you the exact call before it happens, and engineering work arrives as a merge or pull request that no agent is allowed to merge.
faq
which issue tracker does it create work items in?
whichever one the project reaches through an mcp server, such as atlassian for jira or linear's own. it infers the work tracker from the mcp tools the project actually has, asks which to use when more than one could do the job, and reports the gap if none is connected. it never assumes a specific vendor.
will it create tickets without asking me?
no. proposing and then confirming is a hard rule: the backlog planning agent presents the full proposed backlog and waits for your explicit go-ahead before it calls any create tool. it honours the edits and exclusions you make to that proposal.
what does each work item contain?
a title, a description written as acceptance-criteria bullets, a work-item type mapped to what your tracker offers, a best-effort estimate, the parent for child items, and traceability markers including a generated-by-agent label.
what if the tracker rejects a creation?
it retries a failed creation up to three times with backoff and flags anything that keeps failing. afterwards it writes a batch summary to project memory with the spec path, the counts that succeeded and failed, and the identifiers and links of everything it created.
spin up your first project.
one project per client, product, or piece of work, staffed with intake, requirements, build and code review agents.