Empower
Get work done in one keystroke: recognize, compute, search, ask, act across every system. Most everyday answers resolve on-device, at zero tokens.
For the Compliance Officer · DPO
The AI SuperAgent your privacy program can actually stand behind.
Your people are already pasting personal and customer data into consumer AI. Bolt is one SuperAgent that does that same work masked, identity-aware, recorded, and inside a boundary you control, so every AI data flow is one you can explain and prove.
On this page
The crux
Built for the AI age
The exposure you already carry
Governed AI
The shift
One SuperAgent, every system
The governed hub
Why it’s safe to say yes
Governed by design
Policy, not trust
Identity, not a shared bot
Accountability you can produce
Where the data lives
The posture you can attest to
Free, but governed
What you get to point to
Evidence, not assurances
New: the governed browser
The ask · Empower · Govern · Audit
The crux
Every prompt your team sends to consumer AI is a personal-data transfer you can’t see, can’t mask, and can’t prove. Bolt is one AI SuperAgent that masks the identifiers before the model, reaches each system as the user’s own identity, keeps the data in your boundary, and writes what the AI did to a record you can hand an auditor.
Built for the AI age
Empower. Govern. Audit.
The three things an enterprise actually needs from AI at work, in one SuperAgent, not three products.
Govern
PII masked at the model boundary, RBAC over systems and models, identity passthrough, and admin kill-switches on every connector, model and capability.
Audit
What the AI did, recorded and offline-verifiable: answer “what did AI touch?” the instant you’re asked.
The exposure you already carry
Shadow AI is a privacy incident waiting to happen
Consumer AI tools turn ordinary work into ungoverned processing of personal data, with no lawful-basis record and no way to answer for it.
Personal data into a consumer model
Customer names, emails, account numbers: typed into ChatGPT with no masking, no DPA, no record of the transfer.
Cross-border transfer you can’t trace
Once it leaves the device it may be processed and retained anywhere, outside your transfer mechanism, outside your control.
No answer for the DSAR or the regulator
When you’re asked “what did your AI touch, and where did it go?”, today you have no architecture to point to.
Governed AI
Detected sensitive values are tokenized before the prompt leaves your boundary
Use AI on real work without leaking the identifiers in it. Emails, card and account numbers, SSNs, phone numbers and keys are masked before anything reaches the model and restored in the reply, and when you point it at your own LLM, the rest stays inside your boundary too.
privacy
governed
safe
The shift
The same work, but the identifiers never leave
Consumer AI today
Ungoverned processing
- Real names, emails and IDs reach the model
- Personal data leaves the device, retained who-knows-where
- A shared bot reaches more than the user is allowed to see
- No lawful-basis record, no transfer mechanism
- No log to produce when you’re asked to account for it
With Bolt
Masked, identity-aware, accountable
- PII tokenized before the model: it sees «EMAIL», not the person
- Most work answered on-device: it never leaves at all
- Each system reached as the user, respecting that user’s access
- Data stays in your infrastructure: residency is yours to set
- Agent turns, tool calls and approvals in a tamper-evident, offline-verifiable record
One SuperAgent, every system
Connect everything, 1st- and 3rd-party, into one governed surface
Your apps, your models, your tools and agents, even the in-house app only you run, reached as you, governed by IT, audited for everyone.
The governed hub
Bring your systems, your models and your tools, under one policy
Any system, even your in-house apps
Email, docs, tickets, CRM, code, data, and the internal app only your company runs, connected through open MCP, not a fixed catalog you wait on us to extend.
The models you choose
Bring your own LLMs and route by policy. The model is reached for last, on your terms, never by default.
MCP, agents, skills & apps
An open hub your org extends: connect any HTTP MCP server, push the approved tools and agents, and kill-switch the rest.
Identity passthrough
Bolt acts with your identity and respects your access. You see only what you’re allowed to, in every connected system.
RBAC over everything
Admins govern who can reach which system, model, tool and skill, by role, department and group.
Audited & distributed centrally
Roll the hub out to the org, change it once, and keep a tamper-evident record of what the agent did, who approved it, and who was signed in.
Why it’s safe to say yes
You don’t trade control for speed. You get both
Every AI tool promises productivity. Bolt is the one built so the security review passes.
Masked
gated
identity-aware
audited
in your boundary
Governed by design
The six things that make it deployable
PII masked at the model boundary
Emails, SSNs, card and account numbers and keys are tokenized before any model sees them and restored in the answer, and free-form text stays on your own LLM, inside your boundary. The tool gets the real value; detected identifiers are tokenized before the model.
Identity passthrough
Bolt acts with your identity and respects your access, every connected system shows you only what you’re permitted to see.
RBAC over systems, models & tools
Capabilities gated by role, department and group, with admin kill-switches (even proactive AI suggestions can be turned off org-wide).
A tamper-evident audit trail
The agent’s turns, tool calls and approvals are recorded and exportable, a portable proof you can verify offline, built for the auditor in the room.
You define the boundary
No Sparcle server sits in the path. Deploy in your own infrastructure (Helm / Docker) or fully on-device.
The model is reached for last
Most answers come from on-device recognition at zero tokens. The LLM is the exception, not the default, less spend, less exposure.
Policy, not trust
Turn AI behaviours off org-wide, including proactive ones
A real RBAC policy engine decides who can use which AI capability, down to a single one, by role, department and group. Even proactive, background AI suggestions have an admin kill-switch, so “the AI reached out on its own” is never a surprise you have to explain.
rbac
kill-switch
data-minimisation
Accountability you can produce
A record an auditor can verify without trusting us
Every AI action is written to a tamper-evident trail and exported as a portable proof anyone can verify offline. When the question is “show me what your AI did,” you have a file that answers it.
audit
accountability
offline-verifiable
Where the data lives
Residency is yours to control. There’s no detour through us
You can’t govern a transfer you can’t see. Bolt removes the third party from the path entirely.
In your infrastructure
Deploy with Helm or Docker in your own cloud or region, or run fully on-device. You choose where processing happens.
No Sparcle server in the path
Data never reaches our servers in any shipping mode, so there is no transfer to us to account for.
Most answers never reach a model at all
They resolve on-device through local recognition, data-minimisation by default, not by policy alone.
The posture you can attest to
What your privacy program can put in writing
100%
of AI prompts run through masking
structured identifiers tokenized before the model, restored after
Most
answers never reach a model
on-device processing · data-minimisation by default
0
data sent to Sparcle servers
residency stays in your boundary, your region
Free, but governed
Put it in every hand for free. Keep every action under control
The SuperAgent is free to deploy. The governed server is the part you don’t want to skip, it’s the difference between an AI tool and an AI tool your security team can sign off on.
Free client
Productivity for everyone
- Runs on every desktop, no per-seat AI bill
- Hundreds of instant answers with zero tokens
- The whole work surface: recognize, compute, search, ask
- Nothing to lose by rolling it out widely
Governed server
Control your security team will sign off on
- Every AI prompt masked at the boundary
- RBAC over systems, models, tools and skills
- Agent turns, tool calls and approvals in a tamper-evident audit log
- Data stays in your infrastructure, never ours
What you get to point to
Built for your DPO to sign off, not a certificate to take on faith
Bolt provides the controls; your privacy program owns the assessment. Here’s the architecture and the paperwork that make that sign-off straightforward.
Controls that map to your obligations
PII masking, identity passthrough, RBAC, audit and data-locality: the technical and organisational measures you’re asked to evidence.
The paperwork to go with it
A DPA, a DPO playbook, and an employee-notice template, drafted for a corporate-employment lawful basis, ready for your review.
You own the assessment
We don’t claim “certified compliant.” We give you the architecture and controls so your DPO can make the call, and defend it.
Evidence, not assurances
Every claim here is checkable offline.
The audit chain is Ed25519 and Merkle, verifiable with a standalone binary, and the logs carry no PII to leak in their own right.
New: the governed browser
The browser Bolt ships, and the check on the way out
Most AI exposure happens in a browser tab. Bolt ships its own, so what a person types or pastes into a page is screened on the machine before it is allowed to leave it.
A normal browser
Sees the paste, says nothing
- A customer list reaches the AI site intact
- The extension is the only control, and only where it is installed
- No record of what was sent, or why
- Every AI-enabled extension holds its own permissions
Bolt's browser
Names it, holds it, records it
- What was found, grouped by type with a count, without repeating the values back
- The page is held while the person decides, and a written reason is required to proceed
- The reason lands in the audit record as a durable incident, not a counter
- The tab says Guarded only when the guard reported, and Not screened when it cannot
The ask · Empower · Govern · Audit
Give your people an AI SuperAgent you can stand behind.
Roll out the free SuperAgent this quarter. Switch on the governed server and every AI prompt becomes masked, identity-scoped, RBAC-gated, audited, and kept in your boundary, built so your DPO can sign off and your auditor can verify, in a region you control.
GDPR
built for GDPR workflows, DPA with SCCs · Helm / Docker / on-device, your region
1
SuperAgent · one boundary you control end-to-end
0
unaccountable AI transfers
Every other seat
Same platform, argued from where you sit
Each role page makes the case in that role's own language, and links back here.
See it on your own data
Bolt runs inside your perimeter, on the LLM you choose.