Short answers, from setup to billing
The questions that arrive most, organized by topic. When the full answer lives in the docs or the dashboard, the link sits next to the answer.
Getting started
First steps: account, first agent, first content.
How do I create an account?
In the app, at /auth/register: email and password, or Google. The activation arrives by email; once confirmed, the dashboard opens.
Create accountHow much does it cost to start?
Nothing. The Free plan covers one project, asks for no card and does not expire. The paid steps have their prices published at /pricing (Solo US$39, Core US$99, Operations US$199), and contracting starts in a conversation.
Talk to usWhich agents can I connect?
Claude Code, Cursor, Kiro CLI, Hermes Agent, OpenClaw and any MCP-compatible client. Per-agent setup guides live in the docs.
Setup docsHow long does setup take?
Minutes: create the account, mint the MCP token in the dashboard, paste the config into the agent. From then on, hooks start capturing from the first session.
Agents and MCP
How agents talk to the memory.
What is MCP?
Model Context Protocol, the open protocol LLMs use to reach external sources and tools. ValorBrain runs an MCP server at mcpbrain.valor.digital, with OAuth 2.1 and dynamic client registration.
MCP token or API key?
MCP token. It connects the agent to the MCP endpoint and also works on the engine's REST routes, with the same scopes. The dashboard API key exists for one case: pushing a document over HTTP to the ingest route.
Can I use more than one agent at the same time?
Yes, and that is the point of the product: a decision captured in Claude Code shows up for Cursor in the next session. The memory is the same for people and agents.
How do I revoke an agent's access?
In the dashboard, under MCP token settings: revoking the token cuts access immediately (the permission cache expires in under a minute).
How many MCP tools are there?
More than 90 in the current catalog, between canonical tools and aliases, each with a REST mirror. The catalog ships with a parity test against the code.
Memory and permissions
What goes in, how it comes out and who sees what.
How is memory captured?
Through hooks in the cycle of the agents the team already uses: capture happens during the work, with no new screen to open. Whatever needs human validation gets flagged for review.
When something changes, is the old value deleted?
No. The previous version is superseded, with a validity window. Conflicts between sources get detected and quarantined until a person approves or rejects them.
Does search respect permissions?
Yes. Every call runs inside the tenant context (row-level security), and the agent sees what the person using it can see. With source ACL enabled (tenant opt-in), collections from sensitive domains get restricted per person, role or agent, with validity and reason recorded on the grant.
SecurityDo answers come with the source?
Yes: every answer lists the origin document, the date and the collection. When the evidence does not exist, the agent says it found nothing instead of improvising.
Who audits what changed in the memory?
Every mutation records actor, action and before/after state in an append-only trail with a hash chain: editing a row breaks the chain. Every read is recorded too (who read, through which agent and which version), with lookup restricted to administrators.
SecurityPlans and billing
How billing works, what each plan includes and how to scale.
What does the free plan include?
1 person, 1 project and 2 GB of storage. No document quota, nothing billed per response and no card. Enough to prove the product on your real flow.
Do paid plans have a document quota?
No. Documents and searches have no quota and no meter. Each plan's limit is the storage in the table at /pricing (Solo 20 GB, Core 50 GB and Operations 150 GB), and the dashboard warns you when usage reaches 60% and 80%.
How does pricing work?
Per plan and per people: Solo, Core and Operations have a monthly price published at /pricing, with an extra seat at a fixed price. Nothing is billed per response or per document; agents take no seat and searching costs nothing. On the annual cycle you pay 10 months and use 12.
Can I export my data?
Yes. Full JSON export from the dashboard or the API. It is a data-subject right under the LGPD and does not depend on the plan.
What happens if I cancel? Is there a refund?
You can cancel a paid plan at any time. Cancellation takes effect at the end of the period already paid (month or year): access continues until then, and afterwards the account moves to Free with your documents preserved. Cancelling deletes no data; deleting the account erases the data permanently within 30 days of the request. Within 7 days of subscribing to a paid plan, you can withdraw and get the full amount back (right of withdrawal, art. 49 of Brazil's Consumer Protection Code). After that window there is no prorated refund for the unused period, annual plan included. To cancel or withdraw, write to valorbrain@valor.digital.
Security and LGPD
Where data lives, who accesses it and what is not ready yet.
Where is my data stored?
On our own hardware in Santa Catarina, Brazil. Multi-tenant Postgres with row-level isolation; embeddings and reranking run on the same machine's GPU. Self-host and dedicated instances under enterprise contract.
SecurityDoes the engine call external services?
The engine does not touch your sources: the connectors live in the SaaS. No outbound call leaves the engine to fetch third-party content on its own.
Are you certified?
We hold no SOC 2 certification and promise no date. What the security page marks as verified has evidence in code and production configuration; anything unverified shows up as in preparation.
SecurityIs there a DPA and a subprocessor list?
Yes, published on this site, with the data received, the purpose and the location of each third party, including the inference step that leaves the country.
SubprocessorsCommon issues
What to do when something does not work.
My agent gets Unauthorized calling the MCP
The token expired or was revoked. Mint a new one in the dashboard and update the agent's config (for Claude Code, ~/.claude.json).
Search returns zero results
Check in the dashboard whether documents were indexed (collections and counters). Try keyword mode if auto mode fails. If documents are pending embedding, nudge the worker.
I did not get the activation email
Check spam and promotions, add noreply@agents.valor.digital to your contacts, and use the resend option on the login screen. If it persists, write to us.
Is a service down?
The status page shows real availability for the engine and the SaaS. To report, send email with a timestamp and the error message.
StatusThe questions we hear before buying
Common objections, with the answer we give in the conversation.
We already have Notion or a wiki.
A wiki is static. ValorBrain versions, cross-references sources, injects context into the agent, and surfaces a conflict when two sources disagree. The wiki can stay, as one source among others.
Isn't this just RAG?
RAG answers a question. ValorBrain keeps decisions, permissions, versions, and handoffs between agents, and knows when an answer went stale because the rule changed.
Will this be one more tool for the team to open?
No. It lives in the hooks and MCP of the agents the team already uses. Capture happens during work and delivery at question time, with no extra meeting and no new screen.
What about data privacy?
Documents, vectors and the database stay in Brazil. Answer generation uses a model provider outside the country, and the encrypted backup lives in the cloud (Google Drive). In compliance with the LGPD. Tenant isolation, permissions inherited per person and source, read auditing, and no outbound calls from the engine. The connectors live in the SaaS.
What if the agent uses an old version?
Every rule has a version, a validity window, and an owner. The current version arrives before the old one, conflicts between sources and outdated documents get flagged, and material changes go through human review before they circulate. What we promise is traceability: every answer shows which version it came from.
Do we need to rewrite what we already have?
No. Migration happens without rewriting the past: existing documents, decisions, and transcripts come in through ingestion, and versioning starts from there.
How much work does it take to maintain?
Automatic hooks do the capture; memory types (decisions, handoffs, lessons) add structure; the memory health report and auditing show what has aged. Human curation goes where it makes a difference: validation and review.
Why now?
Because every agent that joins without context means starting over again, every time. With one shared memory, each new agent starts from accumulated context, and the leverage compounds instead of diluting.
Did not find what you need?
Write directly to the team. On a business day the answer goes out in under 24 hours; production incidents get a tighter window.
valorbrain@valor.digital