For people building software with AI agents
Your AI agents write the code. Relay makes them show their work.
Relay plugs into Claude Code, Cursor and any agent that speaks MCP. Before an agent touches your project, it has to say what it will build. Before it calls a job done, it has to show what it tested. You see all of it on one board.
Works with the agents you already use, with no per agent fee. $14.99 a month per workspace.
We run our own studio on it: 26 projects, 1,466 commits.
Why it pays for itself
Three things you stop paying for.
- Paying twice for the same work.Two agents can't work on the same thing at once, so nobody's fix gets overwritten and redone.
- Hearing about bugs from your customers.Every finished task comes with what was tested and whether it passed, before anyone sees it.
- Re-explaining your project every morning.Your agents read what they learned last time before they start, so you never brief them twice.
Another tool to learn? No. You keep talking to your agent the way you do today. Relay works in the background, and you look at the board when you want to.
Who it is for
Working alone or as a team, with your agents.
A solo developer
you and your agentsYour problem: A week away and you have lost the thread. So has your agent.
What Relay does: Open the board: what you did last, what your agent did, and what comes next.
A team
a studio, an agency or a company, and its agentsYour problem: People and agents build the same thing twice, and nobody knows who is on what.
What Relay does: Every job is claimed before it starts, and each product or client gets a workspace of its own.
With and without Relay
Four ways agents waste your money. All four happened to us.
Two agents overwrite each other
Without Relay: Two agents edit the same code at once. The last save wins and the other fix is gone.
With Relay: The first agent claims the code. The second sees it is taken and picks something else.
The agent builds the wrong thing
Without Relay: You ask for a small fix. Your agent builds a whole new feature. A week thrown away.
With Relay: Your agent says back what it understood, in one plain line. You agree before any code.
Nobody knows what was tested
Without Relay: Your agent says it is done. It ships. Does everything still work? Nobody wrote it down.
With Relay: The work closes with what was checked: 12 tests run, none failed.
The agent forgets what it learned
Without Relay: This month your agent makes the exact mistake it made last month. It remembered nothing.
With Relay: Before it starts, your agent reads the lesson it wrote last time. The mistake happens once.
Every kind of project
Games, apps, websites: Relay knows what each one needs before launch.
Pick what you're building. Relay lists what usually gets forgotten before launch: saves and updates for a game, uptime and privacy for a website.
Game
A game a player downloads and runs.
How it moves forward
- Prototype build
- Alpha build
- Beta build
- Release build
- Shipped
What Relay makes sure you plan for
- SavesSaves are the player's progress, and there is no other copy of them.
- UpdatesOnce it is installed, it has to be able to update itself.
- DownloadsPlayers download and install it, so the installer is part of the game.
These are starting points, not rules. Add or drop anything, and the board shows what changed.
24projects managed on Relay today
v0.218.1live right now, after 218+ versions since July 2026
137tools your agents can use directly
1,369commits since July 2026
Run on Relay
We sell what we use. These are ours.
Every one of these was planned and checked on Relay, with agents doing the work under the same rules. They're out there, and you can use them today.
This page alone took 0 hours of logged work, on Relay.
UX Handbook
site · live500 short, practical guides to building interfaces, searchable in a second.
Visit ux.tfmstyle.com
How it works
Watch one project go from idea to working app.
A small indie game, built by talking to an AI agent with Relay connected. The agent says how sure it is before it starts. Relay keeps it on plan and writes down every step.
Settings › Connectors
Calendarconnected
+ Add MCP server
NameRelay
Server URLhttps://relay.tfmstyle.com/mcp
API key••••••••••••••••••
Connect
Connecting
✓Relay connected · 137 tools
I want to make a cosy game about restoring an island garden.
B
relay · scope_startnew projectinterview open
What is it for?Calm evenings, growing a garden back.
Who is it for?Cosy game players, on PC first.
What will it not do?No online play yet.
relay · add_projectFerngrovecreated · 5 first tasks
Ferngrove is on your board with its scope written down. First up: fishing.
Start with fishing. Make the bite easier to see.
B
relay · start_workFishing > Biteclaimed
relay · set_confidence88%above the bar
Understood: make the bite easier to see, not a new minigame. On it.
Ferngroveprototype
Fishing > Bite: make the bite easier to seeopen
Map > Night: lanterns on the mapplan Wed
Saves > Autosave: every five minutesplan Thu
Store page > Screenshots: five from the demoopen
Active work
Blue is on Fishing bite
Activity
Blue started Fishing bite
Plan logged · 88% sure
3 commits pushed
Session recordrecording every step
Fishing > Bite: make the bite easier to see
- Read the scope: a calm game, PC first
- Added a splash and a tug on the line
- Decision: keep the old fishing, no new minigame
- Checked it on the slowest computer we support
- Clip attached. Ready for review.
Looks great. Now autosave, so nobody loses their garden.
B
relay · complete_todoFishing > Biteclosed with proof
relay · start_workSaves > Autosaveclaimed
The bite is done and recorded. Starting autosave now.
Ferngroveprototype
Fishing > Bite: make the bite easier to seedone
Saves > Autosave: every five minutesclaimed
Map > Night: lanterns on the mapplan Wed
Store page > Screenshots: five from the demoopen
Active work
Blue is on Autosave
Activity
Fishing bite closed with proof
Blue started Autosave
2 commits pushed
Session recordrecording every step
Saves > Autosave: every five minutes
- Read the lesson from September: load old saves first
- Added autosave every five minutes
- Decision: saves stay on the player's computer
- Proof: 12 save files loaded, 0 failed
- Release note written for 0.3.0
- Every decision, claim and result on the record. Ready for your review.
OverviewFerngrove
Versionnoneno project yet
To-dosnoneno project yet
Records0nothing recorded
Guidelines0 / 6prototype gates
Confidencenonenothing claimed
The board
Know where every project stands, without asking anyone.
Open one board and you know what's done, what's stuck, who has it and what stands between you and the next release. Your agents read and update the same board.
FGFerngroveFG, Ferngrove, alpha laneAlphav0.3.0 to 0.4.0
BoardProductionOverviewSettings
To-dosIssuesPendingIdeas
New to-doTo do5Collapse all
Fishing3
TD-04Fishing > Rods: three rods to unlockadded 2d ago by Mira
TD-05Fishing > Bite: make the bite easier to seeadded 1d ago by Mira
TD-06Fishing > Journal: every fish you caughtadded just now by Theo
Saves2
TD-02Saves > Autosave: every five minutesadded 4d ago by Mira
TD-03Saves > Slots: three save filesadded 4d ago by Theo
FGFerngroveFG, Ferngrove, alpha laneAlphav0.3.0 to 0.4.0
BoardProductionOverviewSettings
To-dosIssuesPendingIdeas
New issueActive1Collapse all
General1
IS-04The game freezes when you open the map at nightupdated 2h ago by Priya
IS-05Crash when a save slot is emptyfrom a crash report · 4 players today
ClosedAllDoneRejectedshowing 1 of 1
IS-03DoneThe garden resets after loading an old saveclosed 3d ago by Mirav0.3.0Reopen
Review3 live, 3 today
SessionsEvaluationArchitecture
Fishing > Rods: three rods to unlockLivenowRoute
Saves > Autosave: every five minutesLivenowRoute
The game freezes when you open the map at nightLivenowRoute
Store page > Screenshots: five from the demoLiveFlagged: waiting on the final logonowRoute
Scheduler5 on the calendar, 1 waiting
AllPlannedDueOverdueOctober 2026
MonTueWedThuFriSatSun
19Fishing bite
20Map freeze fix
21Fishing rodsAutosave
22Fish journal
230.4.0 release
24
25
Unscheduled 1
To-doferngroveFishing > Journal: every fish you caughtSTheoSchedule
Nothing waiting. Every to-do has a day.
FGFerngroveFG, Ferngrove, prototype lanePrototypev0.3.0 to 0.4.0
BoardProductionOverviewSettings
ChangelogGuidelinesScope
Guidelines1/5Prototype build
4 still open in this lane.4 to go before Alpha build
Other1/5
doneCore features implementedPrimary features for the milestone are developed and functionalMiran/a
openNo blocking crashesApp launches and basic workflows complete without critical crashesMiran/a
openKnown issues documentedAll major bugs and limitations are logged in the issue trackerTheon/a
openBuild version assignedVersion number and build identifier follow project standardsMiran/a
openBasic performance verifiedNo obvious performance issues preventing testingTheon/a
Project tracking
See what your project cost, and who did what.
Open any project and see what it cost, who did the work and every step they took, agents included.
Sessions0since 2 September
Tokens099.6% from cache
Output0127K per session
API cost$0every session priced
Time worked0hclaim to close
Every piece of work on Ferngrove is measured: how long it took, what it produced and what it cost.
Who worked on Ferngrove2 people, 2 AI agents
MMira21 to-dos
BBlue41 to-dos
GGreen7 to-dos
PPriya14 issues
People and AI agents side by side, counted the same way.
Blue: make the fishing bite easier to see
- Took itits name on it
- Said it backMira agreed
- How sure91%, and why
- A 3 step planwritten down
- Stoppeda crash came first
- Back to itstep 2 of 3
- Checkedplayed it twice
- Done23 minutes
Every step, in order, including the detour, so anyone can see how it was made.
DocGame design doc, 9 sections
Vault23 notes, linked
Blueprints3 drawn
Work log48 entries
Decisions12 recorded
Lessons6 written
Release notes0.1.0 to 0.3.0
Changelogevery version
The work writes its own documents as it goes. Nothing depends on someone remembering.
Ferngrovealpha
ForPlayers who like slow, cosy games
DoesRestore an overgrown island garden, at your own pace
Will not doOnline multiplayer, until version 2
Made by2 people · 2 AI agents · 31h 20m
NextThe fishing bite, on Thursday
What it is for, always currentWho the project serves and what it will not do, on one page.
How it was madeEvery piece of work, its steps, its cost and what it changed, kept for good.
Who did whatPeople and AI agents side by side, counted the same way.
Why one place
Stop paying for tools that don't talk to each other.
The tracker, the docs, the chat and your agents' memory in one place, so people and agents work from the same picture.
One place to lookEvery task, file, decision and conversation, on one board.
Agents that rememberYour AI works from the same picture your team does.
Fewer subscriptionsStop paying for tools that never talk to each other.
- PlanDecide what to buildto-dos, a calendar, ideas
- WorkPeople and AI agents, one set of rulesagents, who is on what, blueprints
- ReleaseKnow it is readychecks before launch, release notes
- ListenHear what brokebug reports, crash reports, chat
- RememberNever lose the whynotes, decisions, history
The vault
Your AI agents read what the team learned before they start.
Every decision and lesson becomes a note in the vault, linked to the notes around it. Before an agent changes something, it reads the lessons about it, so the same mistake is not made twice.
Before Blue starts on the save code, it reads:
LessonSaves break during updates
DecisionAutosave every five minutes
Lessons5 links
Saves break during updates
In September an update changed how saves are written while a game was saving, and saves broke. Load the old save files before calling any change to saves done.
Linked with
Written by AI agents today
Decision: autosave every five minutes
Lesson: saves break during updates
Read first, every timeAn AI agent starts from what the team already knows, not from an empty chat.
Written as the work happensLessons and decisions are recorded the moment they are learned, not at the end.
Linked, not filedOpen any note and see everything it touches: the project, the decision, the lesson.
Blueprints
See the system before anyone builds it.
Draw it before anyone builds it: how the parts connect, how the work moves, or just the ideas. People draw by hand, agents draw in one go, and everyone works from the same picture.
Somebody presses New workspaceTrigger0 in · 1 outStart
What makes this process begin
Causethe Home page, or a completed sign up
Next
Does the plan allow another workspace?Limit1 in · 2 outRule
A ceiling or a quota
Limitmax_workspaces, from the plans row
Peraccount
Applies to
Ok
Blocked
Mint the opaque workspace idStep1 in · 1 outStep
A unit of work someone does
Ownercontrol plane
Prev
Next
Refused: the plan is fullRejected1 in · 0 outEnd
Deliberately not doing it
Reasonan honest number and the upgrade
Prev
How the parts connect: each box is one piece of the system, wired to what it feeds. This one is real, how a new workspace is created on Relay. People draw it by hand; an agent draws it in one go.
AI agents, with house rules
Watch an agent get caught skipping a step.
- ClaimSay what you are working on before you touch it
- Say it backRepeat the task so a person can correct it
- 80% sureState confidence and risk, or the work waits
- ProofClose with what was checked and what it showed
Built into the tools, not a document.
Nobody has to remember the rules. Skip one and the agent is refused on the spot, and told exactly what it missed. Here is that moment, as it happens.
agent complete_todo(412) relay ✕ Refused: you have not claimed this. Call start_work(412) first. agent start_work(412) relay ✓ Claimed. Payments: retry a failed charge agent set_confidence(86, risk="double charge") relay ✓ 86%. Above the bar, go. agent complete_todo(412, proof="32 tests passed") relay ✓ Closed with proof. In the v0.4.3 notes.
Compare
See what your current tool is missing.
Seven tools, 19 features, each checked against the vendor's own docs. The first group is where Relay is different: agents claim the work, say how sure they are, and prove it before it closes.
| Feature | Relay | Linear | Jira | ClickUp | Asana | Notion | monday.com | GitHub |
|---|---|---|---|---|---|---|---|---|
| Claim before work startsNobody, person or agent, builds what someone else holds | Relay: yes | ?Linear: not verified | Jira: not found in its docs | ClickUp: not found in its docs | Asana: not found in its docs | Notion: not found in its docs | monday.com: not found in its docs | GitHub: not found in its docs |
| Confidence stated first, scored afterHow sure the work will land, checked once it has | Relay: yes | Linear: not found in its docs | Jira: not found in its docs | ClickUp: not found in its docs | Asana: not found in its docs | Notion: not found in its docs | monday.com: not found in its docs | GitHub: not found in its docs |
| Proof required to closeA to-do closes with what was checked | Relay: yes | Linear: not found in its docs | ?Jira: not verified | ClickUp: not found in its docs | Asana: not found in its docs | Notion: not found in its docs | monday.com: not found in its docs | GitHub: partly |
| Every step of a session on recordThe route an agent took, detours included | Relay: yes | Linear: yes | Jira: not found in its docs | ClickUp: not found in its docs | Asana: partly | Notion: partly | ?monday.com: not verified | GitHub: yes |
| Cost of each sessionWhat a session spent, beside what it did | Relay: yes | Linear: partly | Jira: not found in its docs | ClickUp: not found in its docs | Asana: partly | Notion: partly | monday.com: partly | GitHub: yes |
| Official MCP server for agentsAgents read and write the board directly | Relay: yes | Linear: yes | Jira: yes | ClickUp: yes | Asana: yes | Notion: yes | monday.com: yes | GitHub: yes |
| Call transcripts turned into to-dosPoints from a call proposed, reviewed, then filed | Relay: yes | Linear: not found in its docs | add-onJira: with an add-on or a separate product | ClickUp: yes | Asana: not found in its docs | Notion: partly | monday.com: yes | GitHub: not found in its docs |
| Board of to-dos by status | Relay: yes | Linear: yes | Jira: yes | ClickUp: yes | Asana: yes | Notion: yes | monday.com: yes | GitHub: yes |
| Issue and bug tracking | Relay: yes | Linear: yes | Jira: yes | ClickUp: partly | Asana: partly | Notion: partly | monday.com: yes | GitHub: yes |
| Calendar and schedulerPlan days and real deadlines | Relay: yes | add-onLinear: with an add-on or a separate product | ?Jira: not verified | ClickUp: yes | Asana: yes | Notion: yes | monday.com: yes | GitHub: partly |
| Stages and milestones | Relay: yes | Linear: yes | Jira: yes | ClickUp: yes | Asana: yes | Notion: partly | monday.com: yes | GitHub: yes |
| Docs and notes | Relay: yes | Linear: partly | add-onJira: with an add-on or a separate product | ClickUp: yes | Asana: partly | Notion: yes | monday.com: yes | GitHub: partly |
| A graph of linked notesLessons and notes, joined and browsable | Relay: yes | Linear: not found in its docs | Jira: not found in its docs | ClickUp: not found in its docs | Asana: not found in its docs | Notion: partly | monday.com: not found in its docs | GitHub: not found in its docs |
| Diagrams of the systemNode systems, flowcharts, mind maps | Relay: yes | Linear: partly | add-onJira: with an add-on or a separate product | ClickUp: yes | Asana: not found in its docs | ?Notion: not verified | monday.com: partly | GitHub: partly |
| Release notes per version | Relay: yes | Linear: yes | Jira: yes | ClickUp: not found in its docs | Asana: not found in its docs | Notion: not found in its docs | ?monday.com: not verified | GitHub: yes |
| Checks a stage must clearGuidelines gate the next stage | Relay: yes | Linear: not found in its docs | ?Jira: not verified | ClickUp: not found in its docs | Asana: not found in its docs | Notion: not found in its docs | ?monday.com: not verified | GitHub: partly |
| Crash reports on the board | Relay: yes | Linear: not found in its docs | Jira: not found in its docs | ClickUp: not found in its docs | Asana: not found in its docs | Notion: not found in its docs | monday.com: not found in its docs | GitHub: not found in its docs |
| Several workspaces, one accountOne per company or client | Relay: yes | ?Linear: not verified | Jira: yes | ClickUp: yes | Asana: yes | Notion: yes | monday.com: partly | GitHub: yes |
| Free testers and guestsReviewers who file issues at no cost | Relay: yes | Linear: not found in its docs | Jira: yes | ClickUp: partly | Asana: yes | Notion: partly | monday.com: partly | GitHub: yes |
Where every tick comes from
Checked on 21 October 2026 against each vendor's own docs, help centre, pricing page and changelog, on their standard plans. A dash means we searched the vendor's docs and found no such feature, which is not the vendor saying it lacks one. Prices and plan limits change; these are not.
- Linear linear.app/docs/mcp, linear.app/pricing, linear.app/integrations/lincal, linear.app/docs/editor, linear.app/developers/agents, linear.app/developers/agent-interaction, linear.app/docs/releases, linear.app/docs/members-roles
- Jira developer.atlassian.com/cloud/rovo-mcp, support.atlassian.com (usage limits in Atlassian Intelligence; whiteboards; guest access in Jira), atlassian.com/blog/loom, launchnotes.com/release-notes/jira (third party)
- ClickUp help.clickup.com (MCP server; Brain and Super Credits), clickup.com/features/ai-notetaker, clickup.com/blog, clickup.com/pricing
- Asana developers.asana.com/docs/using-asanas-mcp-server, help.asana.com (AI Studio pricing; workspaces), developers.asana.com/reference/getaistudioruns, forum.asana.com (whiteboards)
- Notion developers.notion.com/docs/mcp, notion.com/help/custom-agent-pricing, notion.com/help/ai-meeting-notes, nesslabs.com/notion-backlinks (third party)
- monday.com developer.monday.com (MCP), support.monday.com (AI credits; AI Notetaker), monday.com/helpcenter (WorkCanvas), monday.com/pricing
- GitHub github.blog/changelog (MCP server and Projects, 2025-10-14; usage based Copilot billing, 2026-04-27), docs.github.com (tracking Copilot sessions; Issues), github.com/github/roadmap/issues/372
Workspaces
One workspace per company. As many companies as you run.
Each company you run gets its own board, people, agents and notes, sealed off from the others. One sign in moves you between them.
Northwind Studioone company, one board
owner: youWebsitebeta
Mobile appalpha
Brand refreshprototype
3 projects2 admins5 testers4 agents
Its projects, its people and its agents, on one board that belongs to one company.
PlanBoard and to-dos
PlanScheduler
WorkAI agents and their rules
WorkBlueprints
ReleaseQuality gates and notes
ListenIssues and crash reports
ListenTeam chat
RememberThe vault
Every workspace comes with all of it. Nothing inside is a paid extra.
Owner settingsonly you see this
Invite Theo as an adminsend invite
Invite Priya as a testersend invite
A tester may leave to-dos and issues
A tester may reach the code
Connect an agentnew key
1 workspace, 1 admin$19.98 a month
The owner decides who joins, what each role may do, which agents connect, and pays the bill.
Northwind Studio
Harbor Foods
+ New workspace
A second company gets a workspace of its own. No project, person or note crosses between them.
2 workspaces, 1 admin$34.97 a month
Your workspacesone sign in
Northwind StudioownerEnter
Harbor FoodsownerEnter
Atlas LegaltesterEnter
Harbor Foods2 projects
Ordering > Cart: save a basketclaimed
Menu > Allergens: show on every dishopen
Enter one company and you see only that one. Someone else's workspace can invite you as a tester, for free.
Its own board, stored apartEvery workspace is a separate board. Nothing in one can be seen from another.
The owner holds the keysInvite admins and testers, decide what each role may do, connect agents, and pay for it.
One sign in for all of themA workspace per company, client or team, and you move between them in a click.
Pricing
$14.99 a month per workspace. No per agent fees.
You start with a workspace of your own, and the longer you pay for, the less it costs. Run as many agents as your AI tool allows: Relay adds nothing per agent, and the testers you invite are free.
Monthly
$14.99a month
billed every month
Cancel any month
- 1 workspace
- You, its owner
- Unlimited testers and agents
- Every feature inside
3 months
20% off$14.99a month
billed every month
Cancel any month
- 1 workspace
- You, its owner
- Unlimited testers and agents
- Every feature inside
6 months
30% off$14.99a month
billed every month
Cancel any month
- 1 workspace
- You, its owner
- Unlimited testers and agents
- Every feature inside
1 year
40% off$14.99a month
billed every month
Cancel any month
- 1 workspace
- You, its owner
- Unlimited testers and agents
- Every feature inside
Custom
billed monthlyMore than one workspace, or people to run them with you? Build the plan your company needs.
3
One per product, client or team. You own each one, so you are never an extra.
1
People who run a workspace with you, billed per workspace: the same person on two counts twice.
Always included
Your plan
3 workspaces × $14.99, owner included$44.97
3 admins (1 per workspace) × $4.99$14.97
$59.94a month
Testers and agents never add to this, however many you invite.
Every workspace includes
Subscriptions open soon
- Board, scheduler and scope interview
- AI agents connected over MCP, with checkpoints
- Quality gates, stages and release notes
- Crash reports, issues and team chat
- Notes vault, docs and full history
Need single sign on, a dedicated server or priority help? Talk to us.
Owner
$14.99a month, per workspace
Opens the workspace and pays for it. Creates projects, works on the code and manages everything.
- Can: Create projects and invite people
- Can: Decide what each admin and tester may do
- Can: Change the plan and billing, which only the owner can
- Cannot: Be removed or locked out by anyone
Admin
$4.99a month, on top, per workspace
Runs the workspace alongside the owner, with the same powers unless the owner narrows them. Billed to the owner.
- Can: Create projects and run the work on them
- Can: Invite and remove testers
- Cannot: Change the plan or billing
- Cannot: Change or remove another admin
Tester
Freeas many as you like
Tries the app and leaves to-dos and issues. What a tester may do can be adjusted.
- Can: See the board and report problems
- Can: Leave to-dos, if the owner allows it
- Cannot: Create projects or reach the code
- Cannot: Invite anyone or change a role
Work with us
Want it done for you? We design, build and test it.
Pick one step or all four. We plan it with you, build it in the open and hand over everything. No Relay plan needed.
- Scope first, then the quote.We agree what gets built before any work starts, and the price is for exactly that.
- One price. No surprises.You pay one fixed price for what you asked for, and you get exactly that. Anything new is a new request.
- Not a fit? We will say so.Sometimes the honest answer is no, or not yet. You hear it before you pay, with what would work instead.
Your request
Ask for a quote
Pick the steps you want
Quoted on the scope we agree, never on a guess.
Questions
What people ask before they start.
What does it cost?
Being invited into someone else's workspace as a tester is free. You pay once you open a workspace of your own.
- Your own workspace$14.99 a month, and you are its owner
- Each admin$4.99 a month on that workspace
- AgentsNo per agent fees. Your AI tool bills you as it does today.
- Paying yearly40% off
- Our servicesPriced per step, up to 20% off when you bundle steps
I am not technical. Is this still for me?
Yes. You never touch code to follow along: you read the board like a checklist. If you would rather not run the agents yourself, hire us for the steps you need.
We already use Jira or Linear. Do we have to switch everything?
No. Start with one project and keep the rest where it is. If it earns its place, we help you bring the backlog over.
Does Relay see or keep our code?
Your code stays in your own repository. Relay keeps the work around it: tasks, notes, decisions and the record of every session. Each workspace is stored separately from every other.
Can a client see their project without seeing everything else?
Yes. Give each client their own workspace and invite them to it as a tester. They see their project and nothing of anyone else's, and a tester costs nothing.
Our team uses Cursor or Copilot. Does it still work?
Relay connects through MCP, the open standard that Claude Code, Cursor and GitHub Copilot all support. We run it with Claude Code every day, so that is where it is most polished today.
Do we still need GitHub?
Yes, if you use it today. Relay does not replace where your code lives. Connect a repository and every commit links to its task, so the board shows what is pushed and which task each commit belongs to. No repository yet? Relay works without one.
Does Relay host my app, like Vercel or Replit?
No. Relay keeps the work on plan and on the record while your agents build it. The finished app is yours to ship wherever you like: your own server, Vercel, Replit, an app store or Steam.
If we hire you, do we have to use Relay?
No. We use Relay on our side to document the work. You get the result, whether that is a design, a QA review or a finished build, to use however you like.
Make your agents show their work.
Tell us what you're building. We'll show you Relay on it, or build it with you.
tfmstyle.com/contact