Contextus FAQ
Clear answers about Contextus, plans, and billing.
Contextus is the control plane for governed AI agent workflows. This page explains compile, governed routing, runtime controls, approvals, proof, data handling, and pricing.
AI agent basics
A simple way to understand why agents need context, tools, approvals, memory, and proof.
Read the broader AI agents FAQWhat is an AI agent?
A model call answers once. An agent keeps working toward a goal. It can remember context, call tools, check results, and try the next step. That power is useful, but it also means teams need clear limits around what the agent can see and do.
Why do agents need tools?
A model can write and reason, but a tool lets it act. Tools can search docs, read tickets, write files, call APIs, or run checks. Good agents are built from clear tools with clear names, inputs, outputs, and rules for when each tool should be used.
Why does context matter so much?
Agents make better decisions when they see the right working set: the relevant files, docs, policies, task history, and user intent. A bigger context window helps, but dumping everything into the prompt can still confuse the agent and raise cost. Contextus is built around choosing the useful context first.
What is the safest way to start building agents?
Start by making the workflow work, then make it reliable, then optimize cost and speed. Use stronger models while you are learning the task. Once the behavior is clear, route simpler steps to cheaper models and keep high-risk steps behind policy and approval.
Where do approvals fit in an agent workflow?
Approvals belong at the moment an agent is about to do something risky: write, delete, send, deploy, call the network, or touch credentials. The workflow can pause, ask a human, then resume with the decision recorded.
Product basics
How Contextus helps teams govern context, model choices, actions, approvals, and proof.
What does Contextus do?
Contextus helps teams govern AI agent workflows. It compiles the right context before an agent acts, routes model calls through policy, checks risky runtime actions, requires approvals when needed, evaluates outputs, and keeps audit-ready proof.
How does Compile fit into Contextus?
Compile assembles the right working context before an agent acts. It can include retrieval, ranking, trimming, and compacting as part of the larger control plane for agent context, actions, approvals, evals, and audit trails.
What is SitEmb?
SitEmb is Contextus’s situation-aware context ranking engine. It helps select the files, docs, policies, and task history that matter for the current agent task, while reducing noisy, stale, or unrelated context. In plain terms, it helps agents stay on task instead of drowning in a bloated prompt.
What problem does Contextus solve?
AI agents are moving from chat to action. They can write code, call tools, access systems, and make decisions. Contextus helps teams control what agents see, what they can do, who approves risky steps, and how outcomes are reviewed.
How does Contextus relate to MCP?
MCP gives agents a standard way to connect to tools. Contextus adds the control layer around that connection: compile the right context, classify the requested tool action, ask for approval when needed, and keep proof of what happened.
What is Contextus Gateway?
Contextus Gateway is the governed entry point for model calls. It keeps provider-native request formats intact, makes a policy-aware routing decision, forwards the request to the selected provider, and records routing evidence for later review.
Who is Contextus for?
AI engineering teams shipping agent workflows, platform teams defining policies and approvals, and developer-tool teams embedding agents into IDEs, MCP, CI, or custom environments.
Governance and approvals
How Action Control, policies, and Agent Passports turn agent activity into reviewable decisions.
What is Action Control?
Action Control is the approval and audit layer for agent tool use. It checks requested actions against policy, pauses risky actions, and records decisions in the audit trail.
Where do policy checks help most?
Policy checks help when an agent is about to write files, delete data, send messages, call the network, access credentials, run code, or change configuration. Approvals give the team a clean checkpoint before the action becomes real.
What counts as a risky action?
Examples include writing files, deleting data, sending messages, making network calls, accessing credentials, running code, or changing configuration.
What runtime controls does Contextus check?
At a product level, Contextus can check shell commands, file changes, network access, and tool usage before agents act. Teams can use command approval previews, filesystem boundary checks, network egress policy, runtime action envelopes, and local or MCP agent support without rebuilding their workflows.
Can Contextus block an action?
Yes. Contextus can allow, ask for approval, deny, or redact information depending on policy and risk level.
What is Agent Passport?
Agent Passport identifies which agent is acting, who it acts on behalf of, what it is allowed to do, and when human approval is required.
What is Tool Passport?
Tool Passport answers what tool the agent is using and whether the team should trust it. In buyer-safe language, teams can review trust signals like clean, suspicious, held, quarantined, revoked, or malicious before that tool participates in a governed workflow.
Proof and evals
How Contextus turns approvals, audit logs, and evals into evidence your team can actually review.
What does “prove” mean?
Proof means keeping reviewable evidence: what context was used, which model path was chosen, which action was requested, which policy applied, who approved it, what runtime findings were recorded, and what the evaluation result showed.
Why do evals belong beside approvals?
An approval tells you who allowed an action. An eval tells you whether the workflow produced the kind of result you expected. Keeping them together turns agent work from a vague story into something an engineering or security team can review.
How do proof exports support review?
Proof exports are audit-ready records that support internal review, security workflows, procurement conversations, and incident follow-up.
Can proof include routing evidence?
Yes. When a governed action shares a session with a routing decision, Contextus can include requested model, selected model, provider, policy hash, decision reason, estimated cost, actual cost when available, latency, agent ID, and session ID in the proof bundle.
Runtime modes and lab beta
How Contextus handles normal governed workflows today and more advanced research paths in lab settings.
What is Standard Mode?
Standard Mode is provider-agnostic governance for MCP, editor-native agents, CLIs, SDKs, and managed model workflows. It is the default buyer-facing story: compile context, route under policy, govern runtime actions, approve risk, and keep proof.
What is Hybrid Mode?
Hybrid Mode applies recursive compile planning to managed APIs: summarize, critique missing context, retrieve targeted context, plan, then govern the action. It stays within normal managed workflow boundaries.
What is Recursive Lab?
Recursive Lab is invite-only research for self-hosted or open-weight recursive workers. Lab workflows decode into a human-readable action plan before Runtime Shield, Action Control, and Proof.
Data and privacy
How Contextus handles your code, your data, and how it can be deployed.
Does Contextus see my source code?
For hosted workflows, Contextus processes the context you send through the API or connected tools. For local workflows, teams can keep more processing local depending on setup. The product can be configured based on the team’s security needs.
How is customer data used?
Customer data is processed to provide the service: compile context, route requests, apply policy, record approvals, generate eval evidence, and maintain audit-ready proof.
Can I self-host Contextus?
Self-hosted and private deployment options are planned for larger teams. Contact us to discuss enterprise deployment.
Plans and pricing
What the plans mean, what buyers get at each level, and where billing belongs.
Does Builder include governed routing?
Yes. Builder includes Gateway routing experiments so individual developers can test policy-aware routing decisions before they move into shared team workflows.
What is personal audit history?
Builder includes a personal view of approvals, policy outcomes, and recent workflow activity so individual developers can review what happened without a shared team workspace.
How do Team approvals work?
Team adds shared approval workflows for risky tool actions, plus audit export and role-based access so reviews and escalations belong to the team workflow.
Can I change plans anytime?
Yes. You can upgrade or downgrade as your rollout changes. In-app billing handles plan updates, and pricing changes are prorated when applicable.
Which plan fits a team rollout?
Team brings approvals, governed routing, audit export, and shared governance into one workspace for five people. Builder fits individual validation, while Team supports a coordinated pilot and internal review.
Where do I manage billing after signup?
The Contextus dashboard handles plan changes, invoices, and billing portal actions after signup. The public pricing page helps teams choose the right starting plan.
Which plan should I start with?
Free and Builder are designed for trying the compile and governance loop on personal workflows. Team is the right move when approvals, audit export, role-based access, and shared governance matter.
Plans and billing
Compare plans here. Manage billing in Contextus.
Review the public plan options, then use the signed-in dashboard for plan changes, invoices, and payment details.