BOLTRIG

Put an AI workforce to work without losing control.

Boltrig gives your team a standing group of AI agents that can triage requests, run workflows, update systems and ask for approval when the work matters. You get faster execution without turning trust, security and accountability into manual follow-up.

WORK: DELEGATED
RISK: CONTROLLED
RECORD: COMPLETE

CONTROL

Let agents act, but only inside the rules.

Give agents the tools they need, not blanket access. Boltrig checks who asked, what the agent is allowed to do, whether a human needs to approve it, and whether the request is safe to run before any side effect lands.

ACCESS: SCOPED
APPROVAL: WHEN NEEDED

OPERATIONS

Turn requests into finished work.

SIGNALS.LOGLIVE

A customer issue, release chore, compliance request or internal handoff can become a tracked run with clear ownership, live progress and a result your team can inspect. Agents do the routine work, and people stay in the loop where judgment is required.

REQUESTS: ROUTED
HANDOFFS: VISIBLE

WORKFORCE

Specialists for the jobs teams repeat every day.

Bolt routes work to the right agent, then short-lived workers handle the specific job: release prep, support triage, dependency checks, renewals, notifications and more. Skills are reusable, costs are bounded, and sensitive work can stay on infrastructure you control.

SKILLS: REUSABLE
COST: BOUNDED

VISIBILITY

Know what ran, why it ran, and what changed.

Every agent run leaves a clear trail: the request, the tools used, the approvals requested, the result, and the follow-on work. Leaders get a real operating view instead of a pile of chatbot transcripts and unverifiable summaries.

RUNS: TRACEABLE
EVIDENCE: READY

INTEGRATIONS

Bring your systems, tools and workflows.

Connect the systems your team already uses and package your own domain verbs, workflows and skills. Boltrig can run behind your own product UI, serve agents over MCP, and keep the same governance across every entry point.

PLUGINS: ADOPTED
FRONTENDS: HEADLESS

PRODUCTION

Self-hosted autonomy your security team can accept.

Run Boltrig on your infrastructure, map users through SSO or first-party access, and keep credentials out of agent hands. The value is simple: more work completed by agents, with the controls and evidence a serious organisation needs.

DEPLOY: YOUR CLOUD
PROOF: AUDITABLE

[ REQUEST_ACCESS ]

What_You_Get

AI agents that finish work, with controls your team can defend.

Boltrig is for teams that want agents doing real operational work: triage, updates, renewals, checks, workflows and handoffs. The architecture is there, but the value is simpler: faster work, less manual coordination and a clear record of what changed.

Finish more operational work

Move recurring requests out of chat threads and into tracked agent runs.

One place to delegate. Use Bolt chat, the work board and workflow runs from one console to delegate and track operational work.
Governed agent profiles. Configure specialist workers by runtime, skills, depth, lifecycle and cost tier while caller authority remains the ceiling.
Governed workflows. Define and trigger dependency-ordered workflows whose capability steps still cross the kernel.
Work board. Track agent and human work together, with nested tasks, status and ownership in one view.
Reusable skills. Package the procedures your team repeats, then let agents pull them only when the job matches.

Control what agents can change

Give agents useful access without turning them into unbounded service accounts.

Scoped permissions. Every capability call is checked against organisation, workspace, caller and run authority before it executes.
Human approvals. High-impact actions pause for a human decision instead of relying on a prompt to be careful.
Secrets stay server-side. Credentials are resolved inside Boltrig for one call and are never handed to the agent.
Scoped cost boundaries. Configured organisation and department budgets reserve estimated spend and can hard-stop over-limit spawned agent work.
Sensitive routing. Sensitive work can be routed to local models and blocked from unsuitable endpoints.

Show the evidence

Replace vague AI output with a run record your operators and auditors can inspect.

Live execution stream. Watch reasoning, tool calls, approvals and handoffs as they happen, then reattach if a client drops.
Run inspector. Open any run and walk through the steps, workers and tool receipts behind the result.
Audit and scoped reporting. Governed actions, denials and approvals produce a tamper-evident record, with cost and activity views filtered to the caller's scope.
Knowledge with citations. Keep text, Markdown and PDF originals, search authorised passages and return immutable revision citations while memory stays separate.
Evaluation cases. Run reusable cases under the initiator's grants and inspect assertions, outputs, effective permissions and history.

Meet teams where they work

Use Boltrig's console, your own frontend, or external agent clients.

Console chat. Command the fleet from a live conversation with files, approvals, tool receipts and run recovery.
Headless engine. Build your own UI over Boltrig's HTTP, SSE and MCP surfaces without forking the engine.
MCP in and out. Expose governed capabilities to external agents and consume reviewed MCP tool servers as new verbs.
Signed channel intake. Accept identity-bound Microsoft Teams or generic webhook events after signature, replay and rate-limit checks.
Governed outbound. Send updates through configured outbound webhook endpoints using one approval-gated and audited high-consequence verb.

Bring your domain

Package the nouns, verbs, workflows and rules that make your business specific.

Deployment bundles. Ship project adapters, skills, workflows and manifest policy as a bundle loaded by Boltrig at deploy time.
Live authoring. Author nouns, verbs, bindings, skills, workflows and agent profiles as governed data from the console or API.
External systems. Register reviewed HTTP, OpenAPI, SQL and MCP integrations through the same governed verb model.
Domain guardrails. Compose customer-specific checks from workflow branches and high-consequence capabilities instead of hidden application logic.
No core fork. Keep Boltrig's engine standard while each product brings its own domain package.

Run it where the data lives

Self-host the engine and keep enterprise identity, data and credentials under your control.

Self-hosted stack. Run the kernel, database, console and agent runtime on your own infrastructure.
Organisation and workspace access. Manage members, invitations, workspaces and AI keys with OIDC SSO or first-party invite-only access.
Private credentials. Store model and integration keys server-side, never in browser code or agent prompts.
Production deploys. Build repeatable Docker deployments for the kernel, database, console, workers and durable services.
Tested guarantees. Security and governance claims are pinned to automated tests before they are treated as product guarantees.

[ Request_Access ]Open the console

BOLTRIG

Put an AI workforce to work without losing control.

WORK: DELEGATED
RISK: CONTROLLED
RECORD: COMPLETE

CONTROL

Let agents act, but only inside the rules.

ACCESS: SCOPED
APPROVAL: WHEN NEEDED

OPERATIONS

Turn requests into finished work.

SIGNALS.LOGLIVE

REQUESTS: ROUTED
HANDOFFS: VISIBLE

WORKFORCE

Specialists for the jobs teams repeat every day.

SKILLS: REUSABLE
COST: BOUNDED

VISIBILITY

Know what ran, why it ran, and what changed.

RUNS: TRACEABLE
EVIDENCE: READY

INTEGRATIONS

Bring your systems, tools and workflows.

PLUGINS: ADOPTED
FRONTENDS: HEADLESS

PRODUCTION

Self-hosted autonomy your security team can accept.

DEPLOY: YOUR CLOUD
PROOF: AUDITABLE

[ REQUEST_ACCESS ]

What_You_Get

AI agents that finish work, with controls your team can defend.

Finish more operational work

Move recurring requests out of chat threads and into tracked agent runs.

One place to delegate. Use Bolt chat, the work board and workflow runs from one console to delegate and track operational work.
Governed agent profiles. Configure specialist workers by runtime, skills, depth, lifecycle and cost tier while caller authority remains the ceiling.
Governed workflows. Define and trigger dependency-ordered workflows whose capability steps still cross the kernel.
Work board. Track agent and human work together, with nested tasks, status and ownership in one view.
Reusable skills. Package the procedures your team repeats, then let agents pull them only when the job matches.

Control what agents can change

Give agents useful access without turning them into unbounded service accounts.

Scoped permissions. Every capability call is checked against organisation, workspace, caller and run authority before it executes.
Human approvals. High-impact actions pause for a human decision instead of relying on a prompt to be careful.
Secrets stay server-side. Credentials are resolved inside Boltrig for one call and are never handed to the agent.
Scoped cost boundaries. Configured organisation and department budgets reserve estimated spend and can hard-stop over-limit spawned agent work.
Sensitive routing. Sensitive work can be routed to local models and blocked from unsuitable endpoints.

Show the evidence

Replace vague AI output with a run record your operators and auditors can inspect.

Live execution stream. Watch reasoning, tool calls, approvals and handoffs as they happen, then reattach if a client drops.
Run inspector. Open any run and walk through the steps, workers and tool receipts behind the result.
Audit and scoped reporting. Governed actions, denials and approvals produce a tamper-evident record, with cost and activity views filtered to the caller's scope.
Knowledge with citations. Keep text, Markdown and PDF originals, search authorised passages and return immutable revision citations while memory stays separate.
Evaluation cases. Run reusable cases under the initiator's grants and inspect assertions, outputs, effective permissions and history.

Meet teams where they work

Use Boltrig's console, your own frontend, or external agent clients.

Console chat. Command the fleet from a live conversation with files, approvals, tool receipts and run recovery.
Headless engine. Build your own UI over Boltrig's HTTP, SSE and MCP surfaces without forking the engine.
MCP in and out. Expose governed capabilities to external agents and consume reviewed MCP tool servers as new verbs.
Signed channel intake. Accept identity-bound Microsoft Teams or generic webhook events after signature, replay and rate-limit checks.
Governed outbound. Send updates through configured outbound webhook endpoints using one approval-gated and audited high-consequence verb.

Bring your domain

Package the nouns, verbs, workflows and rules that make your business specific.

Deployment bundles. Ship project adapters, skills, workflows and manifest policy as a bundle loaded by Boltrig at deploy time.
Live authoring. Author nouns, verbs, bindings, skills, workflows and agent profiles as governed data from the console or API.
External systems. Register reviewed HTTP, OpenAPI, SQL and MCP integrations through the same governed verb model.
Domain guardrails. Compose customer-specific checks from workflow branches and high-consequence capabilities instead of hidden application logic.
No core fork. Keep Boltrig's engine standard while each product brings its own domain package.

Run it where the data lives

Self-host the engine and keep enterprise identity, data and credentials under your control.

Self-hosted stack. Run the kernel, database, console and agent runtime on your own infrastructure.
Organisation and workspace access. Manage members, invitations, workspaces and AI keys with OIDC SSO or first-party invite-only access.
Private credentials. Store model and integration keys server-side, never in browser code or agent prompts.
Production deploys. Build repeatable Docker deployments for the kernel, database, console, workers and durable services.
Tested guarantees. Security and governance claims are pinned to automated tests before they are treated as product guarantees.

[ Request_Access ]Open the console

Boltrig - AI agents that finish operational work under control

Put an AI workforce to work without losing control.PutanAIworkforcetoworkwithoutlosingcontrol.

Let agents act, but only inside the rules.Letagentsact,butonlyinsidetherules.

Turn requests into finished work.Turnrequestsintofinishedwork.

Specialists for the jobs teams repeat every day.Specialistsforthejobsteamsrepeateveryday.

Know what ran, why it ran, and what changed.Knowwhatran,whyitran,andwhatchanged.

Bring your systems, tools and workflows.Bringyoursystems,toolsandworkflows.

Self-hosted autonomy your security team can accept.Self-hostedautonomyyoursecurityteamcanaccept.

Finish more operational work

Control what agents can change

Show the evidence

Meet teams where they work

Bring your domain

Run it where the data lives

Boltrig - AI agents that finish operational work under control

Put an AI workforce to work without losing control.PutanAIworkforcetoworkwithoutlosingcontrol.

Let agents act, but only inside the rules.Letagentsact,butonlyinsidetherules.

Turn requests into finished work.Turnrequestsintofinishedwork.

Specialists for the jobs teams repeat every day.Specialistsforthejobsteamsrepeateveryday.

Know what ran, why it ran, and what changed.Knowwhatran,whyitran,andwhatchanged.

Bring your systems, tools and workflows.Bringyoursystems,toolsandworkflows.

Self-hosted autonomy your security team can accept.Self-hostedautonomyyoursecurityteamcanaccept.

Finish more operational work

Control what agents can change

Show the evidence

Meet teams where they work

Bring your domain

Run it where the data lives

Put an AI workforce to work without losing control.

Let agents act, but only inside the rules.

Turn requests into finished work.

Specialists for the jobs teams repeat every day.

Know what ran, why it ran, and what changed.

Bring your systems, tools and workflows.

Self-hosted autonomy your security team can accept.

Put an AI workforce to work without losing control.

Let agents act, but only inside the rules.

Turn requests into finished work.

Specialists for the jobs teams repeat every day.

Know what ran, why it ran, and what changed.

Bring your systems, tools and workflows.

Self-hosted autonomy your security team can accept.