Bring your own model, run agents locally

From a client email to a reviewed pull request

Suporta is the helpdesk that also ships the work. Tickets, sprints, timelogs and KPIs in one multi-tenant workspace, with a fleet of AI agents that plan, code, test, review and document. On your model. Or on your own machine.

company.suportahub.com/ticket/1035-login-returns-500
P1Sprint 14#1035 · 5 points

Login page returns 500 on invalid email

Reported by a client through the Outlook add-in. Elaborated by AI into a developer-ready ticket, estimated at 5 points, and auto-queued to the agent.

Mode
Local agent
Model
Your Claude Code
Token cost
€0.00
Agent pipeline Live
  1. checkPlannerWorked out the approach across 2 repos
  2. checkCoder9 files changed, branch pushed
  3. checkChecksLinting and test suites green
  4. checkReviewerBooted the app, browser checks passed
  5. checkDoc reviewDocs match the change
merge PR #482 approved, ready to merge

Plugs into the stack you already run on

Why Suporta

Intake.
Plan.
Shipped.

Most helpdesks stop at "assigned". Suporta carries the ticket all the way through planning, code, tests, review and documentation, and it keeps every hour, every euro and every pull request attached to the client it belongs to.

smart_toy

Bring your own model, or your own machine

Choose the provider and the model per agent, on your own API key. Or switch a project to Local Agent mode: the work runs on a developer laptop through their own Claude Code subscription, with our MCP wired in. No API tokens billed, no code leaving your git remote.

ClaudeOpenAIGeminiMistralLocal Claude CodeMCP
apartment

Multi-tenant, multi-client

A database per tenant, a subdomain per organisation, and per-company scoping inside it. Your clients see only their own tickets.

euro

Billable by construction

Hours, token spend and agent cost roll up per client and per project, with a markup you set, plus two separate invoice states so nothing is billed twice.

supervisor_account

A human is always in the loop

Agents never merge. They open the pull request, run the checks, and land in a Needs a human queue where you can steer them mid-run, stop them, continue where they left off, or resolve the ticket by hand. Monthly budgets and per-agent switches keep spend bounded.

auto_awesome Six agents, one ticket

Your ticket queue, staffed

Every agent has one job and hands off to the next. They work across all the repositories a project links: one coherent change, one pull request per repo.

  • hub

    Cloud or local execution

    Run agents on your provider key, or queue them to a developer machine. Set it per organisation, override it per project.

  • tune

    A model per agent

    Planner on a thinking model, doc review on a cheap one, review somewhere in between. The catalogue refreshes itself weekly.

  • receipt_long

    Honest, per-client cost

    Tokens are booked per phase, priced per model, and rolled up per client and project with your markup applied for invoicing.

All six agents in detail arrow_forward
graph_3AI dashboard
Live

3

Running

1

Needs you

27

PRs approved

code#1035 Login returns 500Coder, editing AuthControllerlocal
verified#1028 Export misses VATReviewer, browser checkscloud
map#1041 Add SSO to portalPlanner, reading the codebasecloud
Monthly budget38%
One product, five surfaces

Suporta shows up where the work happens

The web app is the source of truth. Everything else, your phone, your inbox, your terminal, is the same ticket seen from where you already are.

Developer docs arrow_forward
terminalBring your own model

Your terminal, your Claude Code, our MCP

Wire Suporta into Claude Code with the MCP server and your tickets become tools you can call: fetch, search, create, update, reply. Then let the worker pool claim queued tickets and run them locally, on your own subscription.

zsh
$ npx suporta-agent init
 registered company.suportahub.com
$ npx suporta-agent start
 claimed #1035 · worktree ready
 claude code session started
 pushed tickets/1035-login-500 · PR opened
worktree per executioncode never leaves your remoteno API tokens billed
desktop_windows

Web app

The full workspace on your own subdomain: tickets, projects, sprints, timelogs, calendar, leave, notes, KPIs and the AI control room.

api

REST API & tokens

Scoped API tokens per developer, per project. Everything the app does is a documented endpoint.

hub

Integrations

Bitbucket, Sentry, Slack, Teams, Microsoft 365 and Google Calendar, two-way where it matters.

What a staffed queue looks like

Measured on the dashboard, per client, every month.

6
AI agents per ticket
5
Places you can work
0
Agent merges without you
€0
Token spend in local mode
How a ticket travels

Four steps, one system

No hand-offs between tools, no copy-pasting a ticket into a chat window, no lost hours.

  1. 1

    It arrives

    From the client portal, an email through the Outlook add-in, or a Sentry issue that becomes a ticket automatically.

  2. 2

    It gets sharp

    AI rewrites a rough client report into a developer-ready ticket, suggests story points, and the planner fills the sprint by real developer capacity.

  3. 3

    It gets built

    Planner, coder and reviewer work across every linked repository, boot the app, run the tests, and approve the pull request.

  4. 4

    It gets billed

    Hours, token cost and agent spend land on the client and project, ready to invoice with your markup.

Built in the open, on ourselves
We run our own agency on Suporta. Every feature on this page exists because a support ticket needed it, including the agents that now close those tickets for us.

Kenneth van den Boogaard

Founder, BitPeak

Put your ticket queue to work

Fourteen days, every feature unlocked, no credit card. Bring your own model, or point it at the Claude Code you already pay for.