For software agencies and consultancies · DACH
Take on the next client project without hiring another senior.
You build software for clients, with your own developers and your own repos. Eight projects in parallel, and in the end everything has to go through the two or three people who know everything. Spedy is the agency software for exactly that case: client boards with budgets, a coding agent on your own runner, and every hour — including its own — on the right client.
- Nordwind Logistik12 · 3 · 2Agent PR waiting for review26/40 h
- Kessler & Partner7 · 2 · 018/30 h
- Vitalis Health9 · 1 · 382 % of budget · notify client41/50 h
- Brandt Industrie4 · 2 · 112/25 h
- Lohmann Retail15 · 4 · 1blocked for 2 days28/45 h
- Sander Mobility6 · 1 · 222/35 h
What we hear
- Everything piles up in front of the same two or three seniors. Every ticket a junior can't finish alone turns into an explanation, a question, a review — and hiring makes the queue longer at first.
- Your developers already work with Claude Code or Cursor. But the agent doesn't know which of your eight clients this is, reports nothing back, and leaves no hour on the project it worked on.
- Eight client projects, eight contexts. Every client has their own board, their own Slack channel, their own docs — and none of them knows anything about the code.
- Status drifts between the pull request and the ticket. "Is that on main yet?" never gets answered out of the PM tool, always out of somebody's head.
- At month end there is no clean record of what went into a client project. Time tracking sits in a second tool, estimate and actual never stand next to each other, and whether an hour was billable is known only to whoever was there.
- Time tracking in the PM tool costs extra. Tempo for Jira, Toggl next to Linear — double entry, and still no number you would want to show the client.
How Spedy solves it
Concrete answers, no marketing promises.
One board per client, one team across them
Every client project gets its own board with its own budget, its own wiki and its own permissions. Your team works across all of them in one workload view: who is on track, who is blocked, what is sitting in review. No tool per client, no context switching between eight systems.
The agent works where the context lives
The ticket hits the trigger status and the job goes to your runner: clone the repo, plan, implement, test, open the PR. The pull request lands back on the ticket as a comment, in your review queue. After "changes requested" it can keep going without a new nudge and fix red pipelines itself. An epic spreads its sub-tickets across a shared branch in waves.
On your machine, with your key
The runner is self-hosted via Docker on an environment you control, with your own Anthropic key or your Claude subscription. Client code never passes through an AI service of ours. That is the difference between "we still need to check that with legal" and a DPA you can pass on to your own client.
Agent hours a human signs off
What the agent books over MCP arrives as a draft in an approval queue — with raw duration and ticket, kept separate from your people's time. It only counts once somebody approves it. You decide whether to bill it to the client before it shows up in any report.
Budget per client board, with carryover
A monthly budget per client, a threshold alert at 80 percent, and billable hours you have not invoiced yet roll into the next period instead of disappearing at the edge of the calendar. Per ticket, the estimate sits next to the actual time. Billable or not is a field you can flip retroactively for a whole board.
The client sees the state, not the stopwatch
The client approves the estimate before work starts, and sees budget and progress at any time — bundled per ticket, without internal comments and without seeing who took how long. If it runs over, you notify them straight from the ticket. Client logins are unlimited and free.
From the field
"We built Spedy because we need it ourselves every day. Every client project sits on its own board with its own budget, the agent works on our runner, and the hour it books is approved by a human. What is written here is what we actually run — it is not a demo setup."
What matters for you
Spedy does more than this. These are the parts that make the difference in your case.
Client boards with monthly budgets
One board per client, budget with carryover and threshold alert, one shared team roster across all of them.
Coding agent on your runner
Ticket in, pull request out — self-hosted via Docker, with your own key. Included on every plan.
Time tracking on the ticket
No second tool. Estimate against actual, billable or not as a field, sync to MOCO, CSV export.
Approval queue for agent hours
What the agent books is a draft first. A human approves before the hour counts anywhere.
Client login included
Budget approval, progress and per-ticket state — without internal comments. Unlimited, free, never counted as a paid user.
GitHub and GitLab on the ticket
Pull requests linked, status moves along, pipeline view across several repos. Spedy does not replace Git.
A wiki per client board
Architecture, onboarding and agreements live where the tickets live — and the agent reads them too.
GDPR, DPA, German hosting
EU hosting in Germany, DPA in the standard contract, passable on under Art. 28 — for when your client asks in an audit.
Our recommendation
Recommended: Pro — €9 per person per month, client logins free
Free is a single seat to look around; that is not enough for an agency, and we would rather say so up front. Pro means unlimited users, the coding agent at no extra cost, and any number of client logins for nothing. At ten people that is €90 a month — at a senior rate of €120 to €160 an hour, at most forty-five minutes. What you pay in LLM costs you pay directly to Anthropic, not to us.
Frequently asked questions
- Our developers already have Claude Code. What does Spedy add?
- Four things an agent in the terminal does not do. It gets the ticket, the client context and the wiki without copy-paste. It reports the pull request back to the ticket, into your review queue instead of somewhere on GitHub. It can keep working after "changes requested" without a new nudge and fix red pipelines itself. And its hours land as a draft in your approval queue, on the right client's board. The bottleneck is not typing — it is context, review and accountability, and that is exactly where Spedy sits.
- How do we get across from Jira, Linear or MOCO?
- You migrate exactly one client project first, not all of them. The importer pulls Jira, YouTrack, GitLab or CSV projects including field and user mapping; statuses come along, comments and attachments with Jira and YouTrack. You keep MOCO: time tracked in Spedy can be synced there per board, and quotes and invoices stay where your accounting already has them. If you like, we will do the first import together on the call.
- Does every client get their own space?
- Every client gets their own board — with its own budget, its own wiki, its own permissions and its own client login pointing exactly there. Your team works across all boards in one roster. What Spedy does not do: a separate subdomain with its own logo per client. The model is boards inside one organisation, not one tenant per client.
- What does it cost to let our clients look in?
- Nothing. Client logins are unlimited and free — only your own people are billed. A client login sees budget, progress and per-ticket state, can approve estimates, and can neither delete tickets nor read internal comments.
- Does Spedy calculate our project margin?
- No, and we do not pretend otherwise. Spedy knows no hourly rates, no margin and no utilisation rate. What you get is clean hours: per client, per ticket, estimate against actual, billable against not, human separated from agent. The valuation in euros still happens in MOCO, Teamleader or your accounting — that is where you export or sync to.
- What do we need for the agent to run?
- Docker on a machine you control, and your own Anthropic key or your Claude subscription. The runner is self-hosted and isolated, with no production secrets; client code never leaves your building. The agent itself is included on every plan, Free included — you only pay your own LLM costs, directly to the provider.
- What do we bill a client for hours no human worked?
- You decide, and you decide before the hour shows up anywhere. Agent time arrives separately and needs a human approval; you can mark it non-billable, show it separately in the client login, or value it with a factor. Spedy does not make that call for you — it just makes it visible instead of letting it vanish into a timesheet.
- Can we get back out?
- Yes. Time exports as CSV at any time, the wiki fully as Markdown, tickets and comments via the REST API. No contract term beyond the month, no export you have to ask us for.
Are you something else?
One client project.Two weeks. Then you decide.
Create a board, import a live project, run the first agent PR. No rollout, no migrating eight clients at once.