Everything your AI needs to get work done
Connect 1,000+ apps. Spinrun loads the right tools when needed, handles authentication, and runs complex work in isolated sandboxes.
Or let Spinrun run the agent for you →- Sequel
- Technics-Group
- Powerod-IT
- Directimo
- LifeLabs
- Notiro
- BirdAI
Claude WorkMatched 38 replies to 21 open deals and logged each one against the right contact. Nine deals have been silent a week or more.
Pipeline is current. Nine follow-ups are drafted and waiting in your outbox.
- 1. Search threads from deal contacts
- 2. Match each thread to its deal
- 3. Draft follow-ups for quiet deals
- ! Pipedrive API tokens are per-user
- ! Follow-ups are drafted, never sent
PIPEDRIVE_UPDATE_DEAL- deals
- 21 updated
- notes
- 38 replies logged
- follow-ups
- 9 drafted
- AGENT
- Claude
- PROVIDER
- Anthropic
- MODEL
- claude-sonnet-5
Bring your AI, or hire ours.
Both run on one gateway. Pick the door that matches how your team works today; the apps, rules and log are shared either way.
Your agent thinks fast.
Give it tools that keep up.
- Fetch open issues
- Rank by severity
- Create board task
- ✓ project slug found
- ▸ board id needed
The right tool, when it matters
Spinrun loads tools as the agent needs them, so the context window stays focused on the work.
- Find tools from the request, not a config file
- Plan multi-step work before execution
- Apply guardrails before the first call
Tools that improve with use
Live calls reveal weak spots. Fixes roll out across the platform instead of waiting for every team to rediscover the same problem.
- Accuracy tuned from live calls
- Optimizations that follow real usage
- A stable API designed for agents
List my open PRs and summarize what needs review.
I need access to your repos to fetch PRs.
Authentication handled for you
Connect an account when the agent needs it. Spinrun handles OAuth and requests only the permissions required for the task.
- Managed OAuth for supported connectors
- Connect accounts inside the workflow
- Keep permission scopes narrow
msgs = []
for ch in run_tool("CHAT_LIST_CHANNELS"):
msgs += run_tool("CHAT_FETCH", channel=ch)
digest = cluster_and_rank(msgs) # 2,183 msgs
run_tool("MAIL_SEND_MESSAGE",
to=me, body=digest) # → 200 SentRun complex work in code
An isolated sandbox handles loops, filtering, and large results, then returns only what the agent needs.
- An isolated runtime for each session
- Compose tool calls from code
- Keep large datasets out of the prompt
Get a deliverable, not a transcript
Ask for a dashboard and get a self-contained page built from connected data, isolated for preview and versioned on every change.
- Interactive HTML and documents, not walls of chat
- Personal by default — widen to a team or the whole workspace
- Revocable public links to share outside the workspace
Controls built into every call
Give each team an endpoint with only the connections it needs. Spinrun checks policy and logs every call. Destructive playground actions wait for approval.
- A scoped MCP endpoint per team, one URL each
- Read, write, destructive — rules enforced at the gateway
- Every call in the activity log, with payload capture you control
| status | last_touch | |
|---|---|---|
| ada@northwind.io | replied | 2h ago |
| li@acme.com | emailed | yesterday |
| sam@globex.co | queued | — |
| ravi@initech.dev | replied | 3d ago |
Every agent gets its own database
An agent that runs every morning needs to remember yesterday. It creates the tables it needs in a private database, and shares some with the rest of the workspace; you browse and export them.
- A private SQLite database per agent — it creates the tables
- Shared workspace tables, so one agent hands work to another
- Browse, page and export on the Data tab; a drop needs an admin
Connectors let your agents act.
Brain gives them memory.
Without Brain, an agent searches one app at a time and may look in the wrong place. Brain indexes your workspace ahead of time, so one search returns ranked answers with citations. Agents add to it too: a report Claude files over MCP, or a Spinrun agent's daily summary, is searchable by the next run.
- One call instead of a dozen per-app round trips
- Semantic and keyword search, fused into one ranking
- Every answer cites the document it came from
- Agents and your MCP client can add documents, so learnings accumulate
One product, every workflow
Point Claude, ChatGPT, or Cursor at one gateway and it can reach every app you use. Ask for the outcome instead of clicking through each app.
Connect each app once through OAuth. Access is scoped and revocable, with no keys to paste into agent settings.
Log this week's replies in Pipedrive
Build a spend summary from Drive
Draft this week's newsletter
Triage new support tickets
Prep me for my 2pm
Send the revenue recap to #sales
Open a PR and tell #engineeringFive lines from model to agent
The model is ready. Give the session access to its tools and let it work.
"mcpServers": {
"spinrun": {
"url": "https://spinrun.ai/mcp"
}
}
Locked down from every angle
- GDPR
A Romanian company under the GDPR and ANSPDCP, with a DPA that binds on acceptance and a countersigned copy on request.
- EU AI Act aligned
Human oversight at the gateway, transparency duties named, no conformity claim we cannot back.
- Core data hosted in Frankfurt, EU
Database, authentication and application functions run in Frankfurt. What is not region-pinned, we say so.
- No training on your data
Neither we nor the model providers we route through may train on your content.
- SCCs for non-EEA transfers
Standard Contractual Clauses, with encryption and access control behind them, wherever data leaves the EEA.