Drop a gateway in front. Traffic runs through policy.
You should not have to rewrite your application to govern it. A Swiftward gateway sits in front of the thing you want to control, and every request passes through your policy on the way.
Your agent does not even know it is governed
This is the point of a gateway. You already have an agent that calls an LLM, an MCP server, or the network. You change one thing, the base URL it points at, or a line in how it is deployed, and every call now runs through your policy. The agent's own code does not change, and does not know Swiftward is there. The alternative is to have your application call our API itself on every decision: more flexible, but your code has to know about us and carry the integration. A gateway asks nothing of your code. Point it at us, and it is governed.
The gateways
An OpenAI- and Anthropic-compatible proxy. Point your SDK at a new base URL and every prompt and response runs through policy. Zero code changes.
Proxies MCP tool calls with per-tool allow and deny, hidden-parameter injection, and a two-axis identity model: who is calling, and what credential goes upstream.
Sits in front of an agent’s outbound calls and controls which destinations it may reach: deny-by-default egress. Destinations today; full traffic inspection built on demand.
Every order on the FIX protocol passes through pre-trade policy before it reaches the venue: position and notional limits, price collars, restricted instruments, per-desk controls.
Takes GitHub and GitLab webhooks, runs the change through policy, and writes back a check or status.
Jail an agent completely
Put the LLM, MCP, and network gateways in front of the same agent and it can reach nothing you have not allowed: which models it may call, which tools it may invoke, and which destinations it may open a connection to. For an autonomous coding agent, or any open-ended agent, that is the difference between hoping it behaves and bounding what it can do. The network gateway controls destinations at the connection level today, with full traffic inspection built on demand; we have run the full jail in a working build.
Identity and keys, handled
The gateway owns authentication, so your agents never hold credentials they should not. Bring your own provider keys, or let Swiftward issue and rotate them, scoped per agent, per team, per user, so the right credential goes upstream for the right caller and nothing broader. On the way through, sensitive data is detected and redacted, so PII, secrets, and proprietary source code never leave your environment inside a prompt. One policy engine decides all of it, and records what it did.
One engine behind all of them
A gateway is just an entry point. Behind every one of them is the same engine, the same versioned policy, the same audit trail. Govern your LLM calls, your agents' tool use, your order flow, and your code changes with one set of rules instead of a separate bolted-together tool for each.