“Dinner for two. One of us is veggie. Nothing spicy.”
Your existing product. A personal agent inside.
Add an agent to your existing interface. Your users keep their accounts; your team chooses its knowledge, permissions, and spending limits.
ONE AGENT PER USEREvery user.
Every user.
Their own agent.
Their history, your knowledge, and the actions you make available.
Connected to their existing account
WHAT YOUR USERS HAVE SHARED
A new chat starts. Their preferences and history are already there.
Past conversationsPreferences & goalsChanges over time
See what carries forward ↗WHAT YOUR BUSINESS KNOWS
Questions about your product get answers from the content your team provides.
Your catalogueYour contentYour policies
Connect your business knowledge ↗WHAT YOUR USERS CAN DO
Your users approve the suggestion. The next step happens right in your product.
Add to a basketSave a planBook a session
Try an action in the chatbox ↗WHAT YOUR TEAM DECIDES
A limit for each person. A shared budget across your users.
Per-user spend capOverall monthly capAlerts & limit behavior
Set the spending boundaries ↗01 / MemoryYour users shouldn’t have to
Your users shouldn’t have to
introduce themselves twice.
Maya opens a new conversation in May. Her coach already knows she’s training for next April, that Saturday is her long run, and about her knee. Your app can pick up with her question.
Your team can follow the conversation, inspect what the agent learned, and correct what it got wrong.
A correction keeps the history.
The old fact is retired and linked to its replacement. A team member’s correction is marked staff, so support can see what the agent believed then and what it knows now.
MARCH · A FIRST CONVERSATION
“I’m training for next April’s marathon. Saturdays are my long run. I need to avoid squats because of my left knee.”
MAY · A BRAND-NEW CONVERSATION
“How should I think about tapering?”
“Your race is next April, so not yet. We’ll build around your Saturday long runs and keep that left knee in mind.”
NEW CONVERSATION. SAME PERSON.
02 / KnowledgeIt answers from
It answers from
the product you actually run.
A programme from your training library. A cancellation answer from your policy. The answer belongs to the product your user is actually in.
Each agent version names the sources it can use. Your coach and your support agent can share a company without sharing a reading list.
↻
Training plan librarySynced catalogue · Coach
312 ITEMS▥
Exercise databaseConnected database · Coach
1,840 ITEMS↗
Support articlesCrawled URLs · Support
94 PAGES↥
Nutrition guidelinesUploaded document · Nutrition guide
PDFIndexed and versioned.Attached to the agents you choose.
03 / ActionsThe next step is
The next step is
right in the conversation.
Your users review the suggestion and tap the action inside your chatbox. Your existing app does the work. The confirmation comes back to the same conversation.
Your TEMPO coachIn your app. Here for you.
●FROM YOUR EARLIER CONVERSATIONS
Next April’s marathon · Tue, Thu, Sat · No squats
YOUR TRAINING PROGRAM
TUEEasy 5k35 min · Conversational pace
THUStrength, no squats25 min · Glutes, calves & core
SATLong run, 14k85 min · The riverside loop
Tap to add all three sessions to your training tab.
✓
You’re all set.Three sessions added to your week. I’ll have them in mind next time we talk.
TEMPO YOUR ACTUAL APP
Training
MON 11TUE 12WED 13THU 14FRI 15SAT 16
TUE
Easy 5k35 min · Keep it conversational
THU
Strength · no squats25 min · Glute bridges, calves & core
SAT · YOUR KEY SESSIONLong run. 14k.85 min · The riverside loop
Your week is waiting.
Add the program to put each session
on the right day.
Add the program to put each session
on the right day.
Scroll inside the training tab to see the full program ↓
04 / SpendDecide what each agent
Decide what each agent
can spend.
Set how much each person’s agent can spend, then set a monthly cap for the environment as a whole. Your team controls both the individual usage and the total.
Choose alerts and what happens at a limit: refuse new requests or switch to a cheaper model.
Explore limits & spend01 / ACROSS YOUR USERSOverall monthly cap
The shared budget for your production environment.
02 / FOR EACH PERSONPer-user spend cap
Each user’s agent works within the limit your team sets.
MemoryActionsYour model
AT A LIMITRefuse requests or use a cheaper model
Your provider’s prices. No markup. ↗TRY THE RULE BEFORE YOUR USERS REACH ITAt the limit,
At the limit,
you decide what happens.
Set the response in your console. Preview it here with one user reaching their monthly model-spend cap.
Set the rules in your consoleLEILA · THIS MONTHLA
78%of this user’s
model-spend cap
model-spend cap
CAP
BEFORE THE LIMITTheir agent can keep helping.
Simulate reaching the cap to preview the policy you selected.
YOUR TEAM’S SIDE OF THE CONVERSATIONA workspace for
A workspace for
the people behind it.
Product authors the agent. Engineering connects your app. Support can trace an answer back to the conversation behind it.
Explore your team’s side of the experience. Inspect an action, open a user’s memory, or adjust a spending limit.
ONE CONVERSATION, ALL THE WAY THROUGHOne conversation.
One conversation.
More context
for the next.
Your users share a preference, find the right option, and take the next step. Their next visit starts with that history already in mind.
REMEMBER THE PERSONDRAW FROM YOUR CONTENTACT INSIDE YOUR PRODUCT
See it in your industryFOUR MONTHS LATER. A NEW CHAT.
Your Piper agentpowered by shardr
Our usual, please?
ALREADY IN CONTEXT
Two mild bowls, one vegetarian.
From your favourite kitchen.
FROM YOUR MENURoast squash + miso salmonNeighbourhood Kitchen · 2 meals
Your basket is ready.Review your meals before checkout.
The conversation changes.
The relationship carries forward.
CONNECT SHARDR TO YOUR PRODUCTFive steps.
Explore the setup in your console
Five steps.
Ready for your users.
Connect the model, shape the agent, and bring in your content and actions. Then add the chatbox to the accounts your users already have.
Your product. Their personal agent.Same account. Same app. Ready to talk.
- YOUR MODEL ACCOUNT
Connect your provider key.
Choose the model your agent uses. Calls stay on your provider account, at your provider’s prices.
- IN THE SHARDR CONSOLE
Give your agent its role.
Set its instructions and tone. Add your catalogue, content, and policies, then publish a version.
- YOUR EXISTING PRODUCT
Connect what your app can do.
Make your existing actions available, then choose which ones the agent can use. Your functions do the work.
- YOUR EXISTING ACCOUNTS
Tell us who’s talking.
Your server identifies the signed-in user. Each person gets their own agent and memory across conversations.
- INSIDE YOUR WEBSITE OR APP
Add your branded chatbox.
Put the conversation inside your existing interface. Test the experience in staging before you launch.
✓Your users can start talking.
For your engineering team
// Your server signs a short-lived token
const token = jwt.sign(
{ sub: user.id, template: "tmpl_...", aud: "shardr" },
process.env.SHARDR_SIGNING_SECRET,
{ expiresIn: "5m" }
)
POST /public/v1/sessions
POST /public/v1/sessions/renew
POST /public/v1/conversations
POST /public/v1/conversations/:id/messages
GET /public/v1/conversations/:id/messagesProduction and staging, isolated.Published versions stay unchanged.Memory staff can read and correct.Records are never hard-deleted.
BUILT FOR YOUR PRODUCT. BRANDED AS YOURS.Make it part