Skip to content

FAQ

Questions, answered plainly.

Every question the site answers, by topic: the program, the product and security.

Design noteThese answers describe how Assent is designed to work. Intent, not results.

Every answer comes from the product documents.

Program

04
Can we use Assent today?

Not yet. Assent is pre-launch: a local prototype, not yet run live. We’re forming a small design-partner cohort to test the first workflow. It starts with conversations about recent changes. Then we reconstruct historical or sandboxed changes with you, partly by hand and under agreed data terms, and compare them with your current workflow and your best improved one.

Source: Validation program

What would a first evaluation involve?

One nominated cross-team API relationship in one GitHub repository, with a versioned OpenAPI definition, identified provider and consumer owners, and a runnable contract or integration test.

Over four to six weeks we’d aim for five to ten eligible real changes, compared with your normal workflow and your best improved one, including your agent vendor’s own orchestration where you have it. People approve decisions and merge throughout.

Source: Validation program

Who pays for agent compute?

Every offer states it up front: either your organization’s own provider accounts, or usage managed by Assent with a visible meter and caps. Personal subscriptions would never be pooled.

Source: Commercial model

What would make you stop?

If teams can’t produce recurring examples; if improved existing tools, including vendor-native orchestration, perform as well; if buyers would rather govern their own agents than have them operated; if execution cost per delivered change can’t be priced sustainably; if authority setup needs continuing expert consulting; or if useful trials don’t lead to paid continuation.

Source: Product thesis

Product

16
Which coding agents does Assent run?

The first managed adapter is planned to be Claude Code, running in an isolated runner that Assent would operate. A second adapter would follow only after it passes control-equivalence tests.

Agents you run yourself, outside Assent, could receive an advisory export. We make no claim of mediated actions, reliable pause or complete audit for those runs.

Source: Authority and enforcement

Will it merge or deploy on its own?

Not in the first evaluation: people approve decisions and merge. Policy-gated merge and release promotion are later steps, enabled per change class only with evidence, explicit customer policy and an authorized policy owner’s adoption. A critical incident would step a class back immediately.

Source: Authority and enforcement

How is this different from our agent vendor’s planning or orchestration?

There’s real overlap, and we won’t pretend otherwise. Our hypothesis is that one continuing chain (the decision, its owner, a scoped grant, the agent work run under it, and the evidence it produced) saves more total effort than orchestration or decision records alone.

The evaluation would compare against your best configured setup. If that does as well, the result should say so.

Source: Product thesis

Questions about decisions.

Does Assent make decisions for us?

No. It is designed to gather evidence and recommend options, labelled as proposals. Accepting a decision would take the authority your organization has confirmed. Agents and AI representatives would never approve.

Source: Decision and assumption register

What if a decision was never written down?

Then the packet would say so. A missing decision that materially affects the change would become a question for its owner; one the evidence can settle would be closed and cited, never invented.

Source: Product workflow

Who counts as an owner?

The people your organization confirms for each concern. Code owners files and service catalogs supply candidates, not proof of who may decide.

Source: Authority and enforcement

Can a decision be approved in chat or by voice?

Not by a message or a spoken yes alone. On every surface, binding approval would need the exact revision on screen and an authenticated, deliberate action. Voice is a later surface, for reading, drafting and consulting.

Source: Authority and enforcement

Questions about grants.

Is the grant a token the agent holds?

No. The contract the runner receives only refers to the grant. The runner authenticates to Assent and checks the grant’s current state before governed actions, so a copied contract authorizes nothing.

Source: Agent execution contract

What happens if a policy changes mid-run?

A grant is designed to stay valid only while its decision, policy, authority, identity, scope and expiry hold. When one changes, dependent work would stop at an enforceable boundary and resume only under a new grant.

Source: Authority and enforcement

Can a grant allow deployment?

Not an implementation grant: production access, external mutation and merge would be outside it. Release promotion would be a later, policy-gated step through your deployment platform.

Source: Authority and enforcement

Who can widen a grant?

No one, by editing it. Wider scope would mean a new decision revision, publication and a new grant. The agent would never be able to widen its own.

Source: Product workflow

Questions about runs.

Can the agent change the checks it is judged by?

No. Protected checks would run from a pinned copy registered outside the workspace. Tests the agent writes would count only as supplemental evidence.

Source: Agent execution contract

What happens when the agent keeps failing?

At the repair limit the stage stops and reaches the attention list with its attempts. Repeated failure that suggests a missing decision is treated as a material unknown, not retried.

Source: Product workflow

Questions about what comes back.

What does “unverified” mean in a report?

Constraints no registered check covers, such as transaction or security semantics. Every report would list them and route them to review; a passing check would never be presented as complete correctness.

Source: Product workflow

What reaches the attention list?

Decisions, reviews, blocked stages and repair escalations that need a person, and nothing else.

Source: Initial prototype and MVP scope

What if a policy changes after approval?

Merge eligibility is withdrawn and the owners are asked to revalidate, even when the code is unchanged.

Source: Worked example: reservation expiry

Security

07
Where does our code go?

We would read only the repositories and paths you select. Selected snippets and contracts would go to agreed snapshot storage and an approved analysis provider. During a run, the code an agent reads would go to its model provider, under the account the execution model names; with your own accounts, your provider’s terms apply.

The provider, region and retention of each would be disclosed. Filtering secrets reduces what reaches a model but can’t guarantee none does, so we would keep the scope small rather than rely on scanning.

Source: Security and procurement

What has to be agreed before you read our code?

In writing: the repositories and paths, storage location, retention and deletion, the model and agent providers and their data terms, the execution model, who can revoke access, and incident contacts. If those aren’t agreed, nothing is read.

Access would start read-only, for discovery and shadow analysis. More permissions would come only with a capability that needs them, such as opening pull requests once managed implementation begins. Synthetic demonstrations need none of it.

Source: Security and procurement

Under whose model accounts do the agents run?

Organization-level provider credentials, to be injected by a proxy outside the sandbox, so keys would never enter the agent’s workspace. Personal credentials and subscriptions would never be used.

Source: Authority and enforcement

Can the runner publish its own required check?

No. A trusted publisher, separate from the runner, would publish it, bound to the candidate and base commits. A passing check is a point-in-time result, so a policy change can still race a merge, and administrators may hold bypass powers; both boundaries must be disclosed and recorded.

Source: Authority and enforcement

What happens if authorization can’t be checked?

The dependent action would be denied, with an explanation of what is missing. Unknown would never be treated as allowed.

Source: Authority and enforcement

Does Assent prevent prompt injection?

We don’t claim to. Content the agent reads can still mislead it. The egress allowlist, held-out keys and review before merge would limit the damage; exfiltration through allowed channels can’t be fully excluded.

Source: Authority and enforcement

Does Assent measure individual developers?

Not individually. An evaluation would record time by person or role so that effort can be counted, but would report it only at team level, and nobody would be ranked.

Source: Metrics and economic evidence

Design partners

Bring one recent cross-team change.

We reconstruct it with you before showing a solution: the decisions, their owners and the effort, agent operation included.