Resources
Get started with Meandr
Guides for every part of Meandr, from your first MCP server to your team's access, and background reading on MCP and agent security.
MCP servers
Bring the tools your agents use: a server at a URL, one on your own machine, or one Meandr runs for you.
Connect an MCP serverAdd a server at a URL and give Meandr its sign-in, from an API key to OAuth, mTLS or AWS SigV4. Then choose which of its tools your agents see.→
Connect an stdio serverRun an MCP server on your own machine. The Meandr CLI dials out, and the server lands in your console, ready for any of your agents.→
Run a cloud MCP serverPick a server from the Docker MCP Catalog or the Official MCP registry, or bring your own Dockerfile, and Meandr runs it on a machine you choose.→
Agents
Connect the agents you already use, or have Meandr run one for you.
Connect an agentGive your agent its workspace URL. Agents that support OAuth just sign in, and others get a key or client ID from the console.→
Run a cloud agentRun Claude Code, Codex or another coding agent on a machine of its own. It keeps working when you close the tab, and you can pick it up from any device.→
Policies
Decide what your agents may do, and when a person has to say yes first.
Your first policyWrite a rule that allows, denies or asks for approval, matched on the agent, the tool and even the argument values.→
MCP human approvalHold an agent's action until an approver enters their one-time code in the chat, and see the approval recorded under their name.→
Policy recipesReady-to-copy policies in plain language: block deletes, require approval over an amount, make a server read-only, rate-limit a workspace.→
Account
Your team, its workspaces, and who can do what in each.
Learn MCP
How MCP works, and what a gateway adds in front of it.
What is an MCP gateway?What MCP is, why you would put a gateway in front of it, and exactly what Meandr terminates and forwards.→
Gateway vs proxy vs routerA decision framework that tells a pass-through proxy from a router and a gateway, with a capability table and honest guidance on when each is enough.→
OAuth 2.1 and PKCE for MCPHow MCP uses OAuth 2.1 and PKCE for agent identity: discovery, dynamic registration, the flow step by step, and the pitfalls.→
The MCP 2026-07-28 specA vendor-neutral reference to the 2026-07-28 MCP specification for anyone running a gateway, from statelessness and header routing to auth hardening and the deprecation clock.→
AWS SigV4 without a proxyWhy AWS-hosted MCP needs SigV4, why naive relays cannot sign it, and how Meandr signs the exact bytes it forwards.→
Agent security
The risks agents bring, and the controls that answer them.
MCP tool poisoningA vendor-neutral field guide: what tool poisoning is, the attack classes, the public data on exposed servers, and concrete defenses.→
OWASP Agentic Top 10Maps the OWASP Top 10 for Agentic Applications to the MCP-layer controls that mitigate each risk, and says where a gateway does not help.→
Argument-level policy vs allowlistsWhy tool-level allowlists fall short against coarse tools, and how matching on the argument values closes the gap.→
What belongs in an audit logWhat each per-call record holds, requests recorded by default, responses opt-in, and how retention scales with your plan.→
Connect your first agent
Run any MCP server in Meandr or connect the ones you already have, then point every agent at one MCP gateway. Policy, approvals and audit cover every call, with no code changes.
Try for freeQuestions? Email us