An AI model should describe an intended operation, not receive a direct handle to the operating system. Nerva’s planned tool gateway is the adapter between those two worlds. It can validate the request, normalize its arguments, send it through permissions, and return a structured result.

Proposal versus execution

A model might produce this proposal:

{
  "tool": "read_file",
  "path": "src/queue.ts",
  "reason": "Inspect the retry policy"
}

The gateway can check that the tool exists, the path is inside the approved workspace, and no extra arguments are hidden. The permission engine then decides whether the read is allowed. Only after those checks does the adapter execute it.

This creates a place to reject malformed paths, unsupported commands, unknown hosts, or an operation that exceeds the task scope. It also gives the UI a clean explanation before approval.

One adapter per capability

File operations, terminal commands, browser actions, and provider requests have different risks. They should not share one generic “do anything” function. A file adapter can enforce workspace paths. A terminal adapter can use an explicit command and working directory. A network adapter can restrict hosts and show the destination.

The gateway should return structured results such as status, output, changed resources, and verification hints. This matters when debugging a backend project: “the command ran” is less useful than “the command ran in this directory, changed no files, and returned exit code 1.”

The gateway is therefore a reliability boundary as much as a security boundary. It keeps model flexibility while making execution predictable enough to review.

The useful design question

The useful question here is not whether an AI agent can produce an impressive demo. It is whether the workflow stays understandable when the task is incomplete, the repository is messy, or the model is wrong. For this part of Nerva, the focus is keeping model output separate from executable tool input. That means each state should have a visible owner, a small input, and a result that another part of the system can inspect.

A concrete scenario

Consider turning a proposed shell command into a typed operation with a visible working directory. A good Nerva run would first state the scope, identify the files or operation it needs, and show the next proposed step. It would keep the original request separate from model-generated instructions. If an action needs approval, the user would see what will happen before it happens. Afterward, the task history would retain the command, result, and verification status instead of reducing everything to a confident paragraph.

This approach also makes a failure useful. A denied permission, unavailable provider, missing package, or changed file is information about the task state. Nerva should preserve it and explain the smallest recovery option. It should not silently broaden the request, retry a side effect, or claim completion because a model supplied plausible-looking text.

What I would test first

For an early implementation, I would test three paths: a normal successful run, a denied action, and an interrupted run resumed after the workspace changes. The key test is whether an adapter can reject an unsafe request before execution. In each path, the interface should make clear what Nerva knows, what it proposed, what actually ran, and what remains unverified.

Useful acceptance checks include:

  • the user can inspect the selected context before model execution;
  • sensitive paths and credentials are excluded or redacted;
  • every side effect has a narrow permission decision;
  • failures remain attached to the step that caused them;
  • a final answer links to evidence instead of implying that unrun work succeeded.

Nerva is still an upcoming project, so these are design and implementation goals rather than claims that a finished product already provides them. That distinction matters: a trustworthy agent should earn confidence through observable behavior, not through a feature list.

The boundary that matters in production

The gateway is also the right place to normalize failures. A timeout is different from a permission denial, invalid arguments, and a command that completed with a non-zero exit code. Keeping those states separate prevents the model from inventing a recovery. A typed gateway contract can be tested with fake adapters before any real shell, browser, or network integration is enabled.

For the first vertical slice, keep the boundary small enough to explain in a code review. Define the input, the decision, the side effect, and the evidence for completion. Add a test for the denied path as well as the successful path. This is where a local-first design earns trust: the user can inspect the context and history on the device, see when a provider is involved, and stop before an action crosses the permission boundary. The design should remain honest about its limits. Nerva is planned software, not a claim that every workflow can be made safe automatically. A clear “cannot verify” state is more useful than a polished but unsupported success message.

A review of this design should use a small fixture repository and record both allowed and rejected outcomes. That keeps the discussion concrete: reviewers can compare the requested scope with the selected context, action, and evidence. It also prevents Nerva from being judged only by how persuasive its response sounds.

The design is informed by the boundaries I have had to think through in CryptoEx and SentinelFi: backend work becomes easier to review when context, permissions, side effects, and verification are explicit.

The control boundary at a glance