Pulls the week's sessions, leads and campaign spend from GA4, HubSpot and Google Ads, compares them with the previous week and sends the report as a PDF to the marketing team.
Specialist agents for your tools and tasks — triggered by ticket, schedule or webhook.
New agent
Search agents …TriggerAllSortNewest
Reset password
Verifies the identity, resets the password in the IdP and securely provisions the new access.
Trigger
Ticket
Active
Create quote
Takes the request details, calculates the quote and generates a PDF for approval.
Trigger
Webhook
Active
License audit
Checks all SaaS licenses, finds those unused for 60 days and revokes & documents them.
Trigger
Schedule · Daily
Active
Post incoming invoice
Reads the invoice, checks it against the purchase order & budget and posts it to the right account in DATEV.
Trigger
Ticket
Active
Weekly marketing report
Pulls KPIs from GA4 & campaign tools, builds the report and distributes it to the team.
Trigger
Schedule · Tue 15:00
Active
Merge CRM duplicates
Checks contacts for duplicates, merges records and logs the history.
Trigger
Webhook
Active
Log sick leave
Reads the notification, records it in the HR tool and informs the team calendar & managers.
Trigger
Ticket
Active
Training reminder
Identifies overdue mandatory training, reminds those affected and updates the HR overview.
Trigger
Schedule · Weekly
Active
Competitor research
Researches competitors on a topic, backs up facts and creates a short dossier.
Trigger
Webhook
Active
InsightsLast 30 days
Analytics
How well your agents perform — success rate, throughput and time saved at a glance.
7 days30 days90 days
Reports
Success rate
96%
130 successful · 5 failed
Tickets handled autonomously
70
▲ +7 vs. previous period
of 84 tickets analyzed
Automation rate
83%
Time saved
38h 38m
Tickets analyzed
84
Follow-up questions
7
in 5% of all agent runs
Avg. agent run
10.8m
Agent runs per day
max 8
85
529.06
7
4
502.07
4
1
105.07
6
7
508.07
4
5
111.07
1
6
814.07
5
7
417.07
1
2
720.07
8
6
523.07
6
1
126.07
7
5
Top 5 agents
Reset password34
License audit30
Post incoming invoice21
Log sick leave18
Create quote9
Cumulative time saved · entire period
62h 37min
June 2026July 2026
Breakdown per agent
Runs, success rate and average duration over the selected period.
AgentRunsSuccessØ DurationTime savedi
Reset password3497%11.5m4h 24m
License audit30100%14.0m10h 00m
Post incoming invoice2195%9.8m4h 00m
Log sick leave1894%7.2m1h 59m
Create quote989%12.3m3h 20m
SystemConfiguration
Settings
All configuration areas in one place.
Instance
General
Instance name, URL, time zone and retention.
System info
Uptime, database and storage of the instance.
Billing & credits
Plan, credits and overage.
Costs
Cost breakdown of agent runs.
Agents
Guardians
A review step before agent results go out.
Ticket pipeline
Mode, Confluence knowledge and routing rules.
Connectors
Jira, Confluence, SMTP and Google Workspace.
AI models
Available models and default model.
Notifications
Email notifications per agent type.
Prompts overview
View and edit all platform prompts.
Agent handoff
Handoffs between agents: mode and limits.
Tools & MCPs
Additional CLI tools and MCP servers for agents.
Files
Files that agents receive in their workspace.
Security
Credentials
Credentials and credential bundles.
Roles & access
Users, roles and SSO.
Chrome
Browser sessions and login profiles.
Audit trail
Change log: who changed what and when.
GDPR-compliantTLS encryptedHosted in Europe
How Taskmate works
This is how Taskmate works.
Taskmate is the platform for AI agents that take over recurring work in your SaaS tools. You define one agent per task: it is triggered by a ticket, a schedule or a call from another system. From there it handles the task autonomously across the tools involved — logging every step through to the finished result.
Inboxes
Request / ticket
Schedule
Webhook
This is how an agent gets triggered.
Processing
Specialist agentrunning
One agent per task — exactly this process, nothing else.
review→decide→run
Instructions (prompt)
Your process in plain words — like onboarding a colleague, with edge cases and limits. This is where your expertise goes.
Build it in the agent builder or straight from Claude Code.
Result
Task closedDone
✓ Task picked up✓ Steps executed✓ Feedback given
Output
Every step is recorded, including screenshots, details and timestamps.
Use cases
Hey Taskmate, please take over.
Whether it is routine work or a one-off task: the Taskmate agent starts your tasks entirely on its own. It clicks through your tools and gets the job done. Whenever something is unclear, it comes back to you to resolve it.
ITrunning
Onboard a new hire
Accounts created in Google Workspace, Jira and HubSpot, roles assigned, welcome email scheduled.
6m 02s · 3 tools · 11 steps
Peoplewaiting
Review a leave request
Handed to the team lead for a decision.
TLFollow-up in the ticket for the team lead
awaiting approval
Licencesdone
Revoke unused licences
9 unused licences released, saving (€870) reported to finance.
Result logged
4m 31s · 0 escalations
Every recurring routine you handle in the browser can become an agent. Below are more use cases.
People
Approve a sick note
Review a leave request
Onboard a new hire
Run an offboarding
IT & access
Reset a password
Set up VPN access
Reset two-factor login
Run a permissions review
Licences
Upgrade a user licence
Run a licence audit
Revoke an unused licence
Support & data
Route a support request
Merge duplicate contacts
Export a report
Sales & contracts
Create a lead
Draft a quote for an enquiry
Prepare an NDA
Renew a contract
Finance
Invoice check
Check expenses
Send a payment reminder
Process a refund
Recurring
Pull a campaign report
Reorder office supplies
Reconcile licence inventory
Verify backups
More routines
Tidy up a project board
Complete time tracking
Coordinate a meeting
And many more …
Raise efficiency, cut costs
Taskmate features
We first built Taskmate for ourselves — to take the load off our own IT team and make hidden costs visible. And we still use it every single day. Every feature here grew out of a real problem; none of it was designed on a drawing board.
Create
Agent builder
Created in conversation, not programmed
You describe in plain English what the agent should take over. Taskmate asks back until the process is unambiguous — and creates the finished agent from it. Nobody writes a script, nobody maintains one.
New agent
Today
Taskmate
Hi! Describe briefly what the new agent should take over — we'll sort out the rest together.
MeMS
Please check daily whether there are new issues for me and handle them as far as you can. Come back to me if anything is unclear.
Taskmate
Sounds like a recurring helper — let's quickly settle the type of agent. Taskmate knows three:
Ticket agent — triggered by an incoming ticket.
Scheduled — runs on a schedule, without a ticket.
On-Demand — called from outside via webhook.
Write a reply …
Connectors
Connected to the systems you already run
Jira, Confluence, Google Workspace, mail server — every connection with its own credentials. Taskmate shows at a glance which connector is ready and where something is still missing.
Atlassian
Jiraready
Confluenceready
Google
Workspaceready
Other
SMTP (email)incomplete
AI models & your own keys (BYOK)
Your model, your key, your endpoint
You define which models are available and which one new tasks use by default. With your own endpoints and stored credentials, Taskmate also runs against a model in your own infrastructure.
Available models
opus ×sonnet ×haiku ×+ add
Default model
claude-sonnet-5
Your own endpoints
IDBase-URLCredential
vllm-internhttps://llm.example.internalvLLM API Token
azure-euhttps://eu.example.internalAzure Key
Operate
Guardian
Recovers failed runs by itself
When a run fails, the guardian examines the cause and decides whether another attempt makes sense at all. You set how often it may retry and for which agent types.
Guardian enabled
Maximum attempts
3
Max. delay
1440 min
inbox_routing
inbox_triage
license_deadline
license_scanner
specialist_dryrun
Logs
Even the failed attempt is visible
Failed runs, warnings and the full transcript of every execution — with code, agent and timestamp. You see not just that something went wrong, but where.
TimeAgentCode
09-18 07:04audit/licencesinvalid_output
09-17 15:20inbox_routingprework_failed
09-17 11:00specialist_dryruncancelled
Improve
Audit
Taskmate audits its own instructions
The audit takes an agent's instructions apart, reports individual findings and proposes a sharper version. Accept it and it becomes the active version — the previous one is kept.
Agent
Licence approval
Status
Running
Created
18 Sep · 09:47
Findings
3 open
Findings
Cost centre is not checked
No threshold amount defined
Reply without a ticket reference
Proposal: new prompt compared with the active one.
Analytics
Where the next agent pays off
Success rate, automation rate, follow-up questions and time saved — per agent and across all runs. So the decision on your next use case rests on numbers.
Success rate · 7 days
97%
28 succeeded · 1 failed
Autonomous
83%
Follow-ups
4
Tickets
76
Avg. runtime
1.0m
Control
Roles & permissions
Every role sees exactly what it needs
Administrator, manager and user come predefined; you create your own roles freely — down to the individual permission. You decide who may create agents and who only watches.
AdministratorSystemall permissions
ManagerSystem24 permissions
Editorial9 permissions
UserSystem2 permissions
Runs8/8
Agents10/10
Credentials3/3
Insights8/8
Integrations even without APIs
Endless integration options
Because Taskmate agents operate your tools' interfaces the way a person does, they can also run routines on platforms that offer no API at all. Whatever you can operate in a browser, Taskmate can take over.
Which use cases do you have? Let's look together at how Taskmate can take over your routine work and save you time and money.
Automation (RPA) has been around longer than AI agents. The difference lies in how Taskmate reaches your tools, what happens when an interface changes — and what can be evidenced afterwards.
Taskmate
Other tools via API
Classic RPA
Works without an API
The agent operates the admin UI like a person — the vendor needs no API at all
Only what the given integration offers
A connector or API is needed per tool
When an interface changes
The agent finds its own way
The integration breaks and needs reworking
The script breaks and needs reworking
Logging
Every step with screenshots and timestamp
A chat history, no evidence of the action
Run logs, usually without screenshots
Language model
Included, or your own key (BYOK)
Tied to the vendor's model
No model involved
GDPR
Processed under EU data protection law
Often processed outside the EU
Depends on the deployment model
Hosted in Europe
Instance in a European data centre — managed or on your own premises
Usually the vendor's cloud
Varies, frequently on-premise
Access and permissions
SSO, roles and permissions, agents isolated in containers
Usually at user level
Through the vendor's platform
Billing
A flat rate based on the effort of the task, regardless of how many users you have
Usually per user, from around €20 per seat and month
Licensed per bot or per runtime
Security
Security and compliance by default
As an international company headquartered in Germany, protecting your data goes without saying for us. That is why we build and run Taskmate to strict European requirements — for the greatest possible security on your side.
GDPR-compliant
All data is processed under EU data protection law — from transfer through storage to deletion.
Hosted in Europe
Your instance runs in a European data centre — either managed by us or entirely in-house on your own infrastructure.
TLS encryption
Data is encrypted in transit. The credentials for your tools are additionally stored encrypted in your instance.
Self-hosted LLM / BYOK
Use the model we include, your own Anthropic key or a self-hosted model.
SSO, RBAC & governance
Sign-in through your identity provider, with no second user directory. You define roles and permissions in detail — down to the individual agent.
Isolated agents
Every agent runs sealed off in its own container on a zero-trust basis. No agent sees another one's data or credentials.
Audit log
Every step is recorded — with screenshot, timestamp and the decision taken. Still traceable months later.
Data retention
You decide how long logs and results are kept. After that they are deleted automatically.
Pricing
Taskmate grows with your routines.
You pay for the effort a task takes, not for the number of people on your team. Start small and extend as more routines come along.
Because Taskmate works independently of APIs, any SaaS tool with a web interface can be connected: the agents operate it in a headless browser the way a person does, so the vendor does not need to provide an interface at all. The tools our customers have Taskmate operate most often include Google Workspace, Atlassian Jira, monday, HubSpot, Billomat and many more.
What is a credit?
A credit is the billing unit for one agent task. Simple tasks (a licence reset, say) consume fewer credits than complex multi-step routines (onboarding across several tools). We deliberately do not bill per ticket, because the complexity behind a ticket varies enormously.
What does “LLM included” mean?
With “LLM included”, Taskmate uses its own language model that we operate for you. You need neither your own API key nor a contract with a model provider. The model runs in European data centres, and your data is not used for training. Processing follows EU data protection law. If you prefer to use your own model, you can do so via BYOK.
What does BYOK mean — and which LLM options do I have?
BYOK (Bring Your Own Key) concerns the language model only, not how Taskmate itself is operated. You can use your own Anthropic key or connect a self-hosted model. That way you decide where the model power comes from.
How is an agent started?
There are three kinds of agents: ticket agents are triggered by an incoming ticket in the connected system, scheduled agents run on a recurring schedule, and on-demand agents are called from outside via an API call or webhook. Each task gets its own agent — for exactly that process, nothing else.
Do I need to know how to code to create an agent?
No. You describe the process in plain English — like onboarding a colleague, with edge cases and limits. Taskmate asks back until the process is unambiguous and creates the finished agent from it. You can do this in the agent builder or straight from Claude Code. Nobody writes a script, nobody maintains one.
What happens when a tool's interface changes?
The agent finds its own way. With classic RPA or API-based integrations, this is usually where the script or the integration breaks and needs reworking.
What happens when an agent gets stuck or a run fails?
Whenever something is unclear, the agent comes back to you to resolve it. When a run fails, the guardian examines the cause and decides whether another attempt makes sense at all — you set how often it may retry and for which agent types. Failed runs stay visible with the full transcript, so you see not just that something went wrong, but where.
How can I trace what an agent did?
Every step is recorded — with screenshot, timestamp and the decision taken, still traceable months later. You decide how long logs and results are kept. After that they are deleted automatically.
Where is my data stored, and is Taskmate GDPR-compliant?
Your instance runs in a European data centre — either managed by us or entirely in-house on your own infrastructure. All data is processed under EU data protection law. Data is encrypted in transit, and the credentials for your tools are additionally stored encrypted in your instance.
Who can create agents?
You decide. Sign-in runs through your identity provider, with no second user directory. Administrator, manager and user come predefined as roles; you create your own roles freely — down to the individual permission.
How much does Taskmate cost?
You pay for the effort a task takes, not for the number of people on your team. Starter is from €500 per month with 1,500 credits, Professional from €1,500 per month with 5,000 credits. Enterprise is available on request — self-hosted, with an included LLM or your own (BYOK). LLM costs depend on the operating model.
We have been building software for 30 years
Taskmate is built by Seibert Group, a software company from Germany operating worldwide: more than 400 employees, 30 years of building modern software — and customers who have stayed just as long.
