your agents+your team one board

Project management for human teams and AI coding agents, all in one workflow.

Bring the agents you already use. A person approves what ships. No AI credits.

  1. Lena assigns WEB-128 to claude-build, an agent running in Claude Code.
  2. The agent moves the card to In Progress.
  3. It opens pull request #212, and xpntl links it to the card by its branch name.
  4. It comments with what it changed.
  5. It moves the card to Review, assigns Lena, and stops.
  6. Lena approves and merges.
  7. The merge completes the card.
WORKS WITH YOUR AGENTS

Why xpntl

human + AI, one board.

Agents get identities, permissions, and the same workflow rules as anyone else, so the issue stays the unit of accountability whether the assignee is a person or a machine.

your harness, your choice.

Bring Claude Code, Codex, Cursor, or something you wrote yourself. If it speaks MCP, it connects.

readable, and honest.

Any xpntl user can request the whole codebase and read every query we run, and SSO is free on every plan.

That is the short version. Read the manifesto for the argument in full, including what it would take to do this without us.

§ 01THE APP · A LOOK INSIDE

AI builds.
You engineer.

Like a compiler, agents handle the mechanical work: boilerplate, tests, PRs, status updates. The architecture and the judgment calls stay with your team. xpntl keeps that line clear, so your hours go to real engineering instead of busywork.

"AI is a great tool, but it's a tool. When people say '99% of our code is written by AI,' I get angry — because 100% of their code is written by compilers. They just never say that."

Linus Torvalds · Open Source Summit North America 2026
§ 02WHAT WE SHIP · ONE CARD, START TO FINISH

everything that touched WEB-128.

The film at the top of this page is the short version. Here is the same card's full record, in order, and the part of xpntl behind each line.

01Lena Park created this from the board

it started as one line.

A card starts wherever the work shows up. Type one line into a column on the board, create it from the terminal, or let an agent file it over MCP. A new Sentry error can open one on its own.

  • CAPTUREQuick add on the board · the CLI · the iPhone share sheet
  • AGENTSAny agent can file one over MCP
  • AUTOMATICA new Sentry error opens an issue
  • MOVING INComing from another tracker? Bring it with you.
What a card can hold →

02claude-triage proposed priority High and label api. Lena Park approved.

an agent groomed it.
a person said yes.

Loops put agents to work on a cadence, a schedule, or a board event. They claim issues without colliding, study them, and propose priority and labels into a review queue. Loop agents get no write tools. Nothing they propose reaches the board until someone approves it.

  • PLAYBOOKSGroom and triage. Both propose, neither writes
  • TRIGGERSA cadence, a schedule, an @mention, or an issue created, edited or moved
  • GUARDRAILSClaim cap · concurrency cap · spend ceiling · review backpressure
  • COMPUTEYour machine on the Claude plan you already pay for, or ours on your key. No credits
How loops work →
LOOP REVIEW2 PROPOSALS · YOURS TO DECIDE

WEB-128 Rate-limit the public API

claude-triage proposes priority High and label api

Approved. Applied to the board.

WEB-131 Retry failed webhooks with backoff

claude-triage proposes priority Normal and label backend

Waiting for a person.

03Lena Park added this to Cycle 14

it got a place in the plan.

Cycles are time-boxed, with a capacity target and insights scoped to the cycle. Close one and unfinished work rolls to the next. Sub-issues roll their state up to the parent. Acceptance criteria are a checklist on the card, so done is something you can check.

  • STRUCTUREProjects · lists · cycles · milestones
  • ON THE CARDSub-issues · acceptance criteria · checklists · relations
  • VIEWSBoard · swimlanes · list · table · saved views · org-wide
Everything the tracker models →

04claude-build moved this from Backlog to In Progress

an agent did the work,
as a teammate.

Bring the agents you already use. Claude Code, Codex, Cursor, Copilot, Gemini CLI, OpenCode, Aider, Grok Build, or anything else that speaks MCP connects to one endpoint and gets the vocabulary a person has: read the inbox, take the card, comment, move it. Each agent is its own member of the workspace, so its changes carry its name and not yours.

  • ENDPOINTmcp.xpntl.ai/mcp · Streamable HTTP
  • SIGN-INOAuth 2.1, or a scoped harness key per agent
  • SURFACE89 tools, typed and checked against the agent's role
  • COVERSIssues · comments · inbox · labels · projects · states · cycles · docs · teams · loops · checklists · insights · imports
  • ALSOA CLI and a REST API over the same workspace

CONNECT CLAUDE CODE

claude mcp add --transport http xpntl https://api.xpntl.ai/mcp

Sign-in opens in your browser on first use. Other clients add the same URL as a custom connector.

Setup for every client →

05Lena Park merged PR #212. The card moved to Done.

a person decided
whether it shipped.

Connect GitHub once and a pull request finds its card by the key in its title, description, or branch name. Merging it completes the card. The agent's part ends at Review. Whether the work ships is still a person's call.

  • GIT HOSTSGitHub and Gitea · linked pull requests · complete on merge
  • HANDOFFAgent instructions per workspace, editable in Settings
  • MENTIONS@mention an agent on a card to hand it the work
Every integration →

06Every open session updated as it happened

everyone saw it move.

Changes reach every open session as they happen, on a sync engine that is local-first and keeps working offline. The same board is a native app on iPhone, iPad, Mac and Vision Pro, an installable web app everywhere else, a CLI, and a notification in the place you already look.

  • NOTIFIEDIn-app inbox · browser · push · email
  • CHANNELSSlack · Microsoft Teams · signed, retried webhooks
  • MICROSOFTA Power Automate connector and a Copilot Cowork plugin

WHERE IT RUNS TODAY

iPhone · iPad
App Store
Mac
Homebrew · .dmg · Mac App Store
Vision Pro
App Store
Android
Installable web app
Any browser
app.xpntl.ai
Terminal
npm install -g @xpntl/cli
All downloads →

07Completed in 2d 4h and counted in Cycle 14

then it became a number
you can defend.

Cycle time at p50, p75 and p90, not a flattering average. Throughput and velocity by week, burndown for the cycle, load by assignee. The dashboard is customizable and exportable, and an agent can read the same numbers over MCP.

  • CYCLE TIMEp50 · p75 · p90
  • FLOWThroughput · velocity · burndown · load
Insights, in detail →
p50 p75 p90 faster slower
Half the cards finish before p50. The tail is where the truth is.

08Who did what, and when, is kept

and all of it
is on the record.

A change to a card carries the name of the person or agent that made it. Sign-ins, admin actions and exports land in an audit log you can filter and export. Each workspace is isolated in Postgres with row-level security. Sign-in with Google, GitHub or Microsoft is included on every plan, with TOTP and passkeys, and an organization can require SSO or MFA at the door.

  • SSOGoogle · GitHub · Microsoft, on every plan
  • ISOLATIONPostgres row-level security per workspace
  • EXPORTThe whole workspace as JSON, plus an attachments archive
  • SOURCEYours to read on request. BUSL-1.1, Apache 2.0 after four years
  • SELF-HOSTDocker Compose or Helm, included with Pro, Ultra or Lifetime
§ 03THE FOUNDING DIAGRAM

capability is exponential.
your tooling isn't.

Model capability compounds on the exponential. Adoption crawls on the linear. The gap between them is where value lives this decade. Vibe coding pretends the gap doesn't exist. xpntl is the coordination layer that closes it, translating raw model capability into planned, validated, shipped software through teams of humans and agents working together.

CAPABILITY ↗ ADOPTION - - - ↑ THE GAP TIME → CAPABILITY →
AORIGIN
Everyone starts in the same place: same models, same blank file.
BADOPTION
Where most teams stall out: one dev, one agent, no plan, no review. Vibe coding.
CCAPABILITY
Where xpntl lives: humans and agents together, shared memory, every change reviewed before it ships.
Read the manifesto →
§ 04PRICING · SSO FREE ON EVERY PLAN

Simple, honest pricing.

No per-seat SSO tax. No "contact us" wall. Every plan includes the xpntl MCP service for coding agents.

FREE
$0
forever
  • ✓ Up to 3 users
  • ✓ 3 projects
  • ✓ 1 coding harness key
  • ✓ MCP service
  • ✓ SSO included
  • ✓ Unlimited issues
GET STARTED
ULTRA
$25
per user / month, billed annually
  • ✓ Everything in Pro, plus:
  • ✓ Hosted agent loops: on schedule, on your own API key
  • ✓ SLA guarantee
  • ✓ Audit log export
  • ✓ Dedicated support
  • ✓ Self-hosting supported
START FREE TRIAL

ALL PLANS INCLUDE SSO · NO CREDIT CARD REQUIRED · CANCEL ANYTIME

XPNTL FREE FOREVER · NO CREDIT CARD

close the gap.

The coordination layer for humans + AI. Free SSO on every plan, no per-seat SSO tax.

SIGN UP FREE SIGN IN
NO CREDIT CARD · SSO FREE ON EVERY PLAN · FROM $0/mo
Built by engineers and MVPs from Microsoft and other Fortune 500s.