Admins decide which tools your team's AI may use and which calls an admin must approve. Weave writes the tool calls to the audit log. Operated in Germany.
Works with Claude Desktop, Cursor, Windsurf, VS Code Copilot, ChatGPT Desktop, Cline, Continue, Zed, and any MCP-compatible tool.
Your AI clients speak MCP. Your tools expect API calls. Weave is the gateway, with encryption, rules and an audit log of the tool calls.
One admin console for your team members and your AI agents — same identities, same tool assignments, same live monitoring, same audit trail. The unified surface most companies stitch together from five SaaS tools.
See Team & Agent GovernanceTool calls in a hash-chained audit log, approval rules for sensitive tools, a key per organization and data centers in Germany.
Talk to UsConnecting an AI to your tools is the easy part. Making sure it can't accidentally email the wrong customer, leak a credential, or run a destructive command — that's the part everyone ignores. We built four control layers so you don't have to.
Ten scanners check tool calls before execution for prompt injection, data exfiltration, destructive commands, secrets and more. Depending on the policy, Weave blocks a detection or only logs it.
Admins set approval rules per tool pattern, server and who calls, person or agent. When a rule matches, the admin sees the exact arguments and decides in the dashboard; the client can then retry the call.
Credentials are sealed with a per-tenant DEK, itself encrypted by the platform KEK. A database dump alone reveals nothing. Keys rotate per organization, so a compromise in one tenant has zero blast radius outside it.
Some tools should never leave your machine. Run file systems, local databases, and internal-network services through our Bridge Agent on the user's laptop. The cloud control plane stays in sync; the data never leaves the device.
Add tools from the catalog, store credentials once, encrypted, and decide what the AI may do.
Each member pastes one gateway URL into Claude, Cursor or another MCP client and works with the team's tools.
An admin decides calls covered by an approval rule in the dashboard. The audit log shows who called which tool and when.
API keys sit in private config files, each person decides alone what their AI may do, and when something goes wrong nobody knows which call it was.
Weave runs on servers in German data centers (Hetzner, see the DPA). Which data goes to external services is up to the tools you choose.
Fully compliant with GDPR. DPA (Art. 28) available. Your privacy is non-negotiable.
All credentials encrypted with AES + HMAC. Per-tenant encryption keys for cryptographic isolation.
Per-user containers and per-tenant encryption keys guarantee complete data separation between organizations.
See what changes when your AI capabilities are managed, not scattered.
Docker setup per tool, manual config files
One-click install from the marketplace
API keys in .env files, no encryption
Credential vault with envelope encryption
Separate URL per tool, re-configure each client
One Universal Gateway URL, all clients
No visibility into what AI does with tools
Hash-chained audit log of tool calls
Per-user chaos, no team governance
Central admin, shared rollout, access policies
Read our Privacy Policy and the DPA (Art. 28 GDPR) for all details.
Free during the beta, no credit card needed. If you sign up on your own, you get a team of one.