How can we help?

Go to /auth/register. Use email + password (or Google OAuth). Get an activation email. Once confirmed, /dashboard is accessible.

Nothing. The free plan is free forever, no card: 1 person, 1,000 documents, unlimited memory search and 50 sourced answers per month. To see the whole product, the top self-serve plan (Empresa) is unlocked for 14 days, also without a card.

5 minutes total: signup (1min), connect first agent via MCP (2min), test query (2min). The agent setup wizard guides each step.

Anything that speaks the Model Context Protocol (MCP). Tested in production: Claude Code, Cursor, Kiro CLI, Hermes Agent, OpenClaw. The list grows as the protocol gets adopted.

Add ValorBrain MCP server in claude_desktop_config.json: { "mcpServers": { "valorbrain": { "url": "https://mcpbrain.valor.digital", "auth": "oauth" } } }. Claude does OAuth automatically when first calling a tool.

No. ValorBrain uses OAuth 2.1 with Dynamic Client Registration. Each agent registers itself, you authorize on the dashboard, no keys to copy/paste.

Cursor needs MCP enabled in settings (early-2026 feature). Confirm Settings → Features → MCP is on. If still not, our MCP server is at https://mcpbrain.valor.digital. you can use REST direct via /api/v1/* meanwhile.

With documents indexed, open Ask (/knowledge/ask) or use Ask in the main nav. Type a plain-language question; the answer lists sources, origin, and score. That is the product aha moment.

On /knowledge (Find tab) use the search bar. Keyword + semantic by default. Filters: collection, content type, dates. Prefer Ask for natural-language questions with citations.

Lifecycle hooks (UserPromptSubmit, Stop) listen to your session. The decision-extractor runs a classifier that pulls structured observations: decisions, problems, milestones, antipatterns. Saved as documents with content_type set automatically.

Yes. Two ways: (1) bulk import via the wizard at /knowledge → "Import" button (Excel/CSV), or (2) automatic via watcher: install the local CLI, point at a folder, every .md file syncs in real time.

OAuth connectors for Notion, GitHub, Slack, Drive. Webhook /api/v1/ingest accepts arbitrary JSON. Watcher CLI for local files. Schema packs adapt the decision-extractor per domain (coding, marketing, legal, ops, HR, support).

Hybrid: BM25 (keyword) + dense vector (semantic) + RRF fusion + cross-encoder reranking. Auto-routing for "why" / "what changed" / "decision" detects intent and weights subsystems.

You pay for people and for AI answers. Documents, memory and search are unlimited on every paid plan: no document quota, no search meter. The full ladder with each plan price is on /en/pricing.

The top self-serve plan (Empresa) is unlocked for 14 days with no credit card: every limit, every connector. When the trial ends the account drops back to the free plan and your data stays — nothing is deleted.

A plain-language question answered with the source document cited (Ask, at /knowledge/ask). Only the generated answer counts: searching and browsing the memory consumes nothing, because there is no LLM cost. Each plan includes a monthly allowance; beyond it we charge per answer at the price listed on /en/pricing.

Documents and search, on paid plans — literally, no quota. There is a fair-use storage ceiling (GB per plan) that in practice is worth more than a hundred thousand documents; it exists for the pathological case, not the normal customer, and if you get close we call you before anything else. The free plan is the only one with a document cap.

Settings → Users → Invite. Each plan already includes a number of people; above that, an extra person costs a fixed monthly price (cheaper on Empresa than on Time), listed on /en/pricing. Sharing documents with the team costs nothing.

30-day guarantee: if you do not like it, we refund what you paid. On annual billing you pay 12 months for the price of 10.

Yes. Settings → Subscription → Cancel plan. Access continues until the end of the period you already paid for; after that the account drops to the free plan and your documents stay where they are.

Default in Brazil (LGPD). Postgres + pgvector with RLS (Row-Level Security) enforcing tenant isolation: no app-side query can cross tenant boundary. Enterprise plan has on-prem option (your infra, zero third parties).

OAuth tokens, BYOK credentials and API keys are encrypted at rest with AES-256-GCM (key on the server, not derived from password). DB only sees ciphertext + IV + auth tag.

Only members of your tenant. Granular roles (admin, member, viewer) scope what each one can do. RLS enforced at DB level. even a SQL injection in app code can't see other tenants' data.

In progress. Compliance roadmap public at /en/security. LGPD already enforced (default region Brazil). HIPAA-ready on Enterprise plan.

Settings → Users & Permissions → "Invite". Send email with link. Member account created on confirm + scoped tenant access. People included: free plan 1, Time 10, Empresa 50, Enterprise unlimited — extra people cost a fixed monthly price on the paid plans.

Settings → Audit logs. Every API/tool call logged with user_id, action, timestamp, status. Filters by user, period, action type. Export to CSV for compliance.

Standard roles for now (admin/member/viewer). Custom roles with granular scopes on the roadmap (Q3 2026). For special workflows, Enterprise plan covers.

1. Check /status · ValorBrain might be in maintenance. 2. Confirm OAuth in /settings/mcp-tokens still active. 3. Re-authorize: remove the agent's tool call permissions and re-authorize on the dashboard. 4. If persistent, contact support.

First: check /status (degraded indicates load). Second: long queries (>200 chars) trigger expansion + reranking. Try simpler queries or use intent_search direct ("why did we decide X") which routes faster.

Soft-delete by default: doc is archived, not destroyed. Settings → Knowledge → Archived → "Restore". After 30 days the actual purge runs.

Email [email protected] with: error description, expected behavior, screenshots/logs if any, the docid in case it's related to specific data. We respond in <24h on weekdays.