# Vera.AI

Vera is an outbound operator for evidence-grounded prospecting, drafting, sending, reply handling, and follow-up.

Canonical product URL: https://vera.symph.ai
API base URL: https://api-vera.symph.ai

Human and agent docs:
- /docs
- /docs/journeys
- /docs/quickstart
- /docs/context
- /docs/prospects
- /docs/outbound
- /docs/replies
- /docs/runs
- /docs/autonomy
- /docs/authentication
- /docs/api-keys
- /docs/reference

External agent journey:
1. Connect: use a bearer API key, read /v1/workspace, verify Gmail status, verify sender aliases, and check project readiness.
2. Build seller context: create or select a project, upload notes, URLs, and files, then run discover_business.
3. Review knowledge: read /v1/projects/{project_id}/knowledge and only rely on approved, evidence-backed claims.
4. Shape prospects: run search_prospects, run research_prospects, inspect prospect research, and approve or revise research.
5. Create strategy: run generate_strategies or approve research so Vera queues a single generate_strategy child run.
6. Draft outreach: run generate_drafts or approve strategy so Vera queues a single generate_draft child run.
7. Send only approved drafts: read drafts, confirm review_status and lifecycle_status, verify Gmail sender, then call draft:send with confirm_reviewed true.
8. Handle replies: sync inbox, inspect inbound messages, generate reply drafts, mark takeover when a human should handle the thread.
9. Process follow-ups: read schedules and tick due follow-ups. Production should use Cloud Scheduler for internal ticks.
10. Continue safely: call agent:continue to ask Vera for the next safe preparation step or blockers.

Public run operations:
- discover_business
- search_prospects
- research_prospects
- generate_strategies
- generate_drafts
- generate_reply_draft
- generate_follow_up_draft

Review child operations that may appear after decisions or continuation:
- research_prospect
- generate_strategy
- generate_draft

Primary API paths:
- GET /v1/workspace
- GET /v1/projects
- POST /v1/projects
- PATCH /v1/projects/{project_id}/policy
- GET /v1/projects/{project_id}/readiness
- GET /v1/api-keys
- POST /v1/api-keys
- POST /v1/api-keys/{api_key_id}:revoke
- DELETE /v1/api-keys/{api_key_id}
- GET /v1/integrations/gmail/status
- GET /v1/integrations/gmail/aliases
- PATCH /v1/integrations/gmail/selected-alias
- POST /v1/projects/{project_id}/context
- GET /v1/projects/{project_id}/knowledge
- PATCH /v1/knowledge-claims/{claim_id}
- DELETE /v1/knowledge-claims/{claim_id}
- POST /v1/projects/{project_id}/runs
- GET /v1/runs/{run_id}
- GET /v1/projects/{project_id}/runs
- GET /v1/projects/{project_id}/next-actions
- POST /v1/projects/{project_id}/review-decisions
- POST /v1/projects/{project_id}/agent:continue
- GET /v1/projects/{project_id}/prospects
- POST /v1/projects/{project_id}/prospects
- PATCH /v1/prospects/{prospect_id}
- GET /v1/projects/{project_id}/prospect-research
- GET /v1/projects/{project_id}/strategies
- GET /v1/projects/{project_id}/drafts
- PATCH /v1/projects/{project_id}/drafts/{draft_id}
- POST /v1/projects/{project_id}/drafts/{draft_id}:send
- POST /v1/projects/{project_id}/inbox:sync
- GET /v1/projects/{project_id}/email-messages
- POST /v1/projects/{project_id}/email-messages/{message_id}:takeover
- POST /v1/projects/{project_id}/inbound-candidates:sync
- GET /v1/projects/{project_id}/inbound-candidates
- POST /v1/projects/{project_id}/inbound-candidates/{candidate_id}:accept
- POST /v1/projects/{project_id}/inbound-candidates/{candidate_id}:ignore
- GET /v1/projects/{project_id}/follow-up-schedules
- POST /v1/projects/{project_id}/follow-ups:tick

Agent rules:
- Do not invent claims, metrics, customer proof, integrations, prices, contact details, availability, or prospect pain.
- Prefer approved knowledge claims and source evidence.
- Preserve review gates unless the project is explicitly autonomous.
- Stop when Gmail is disconnected, the sender is unverified, readiness has blockers, or a draft is not ready_to_send.
- Stop for pricing asks, objections, unsubscribe intent, unclear replies, human takeover, risky claims, and sensitive decisions.
- Read the journey resource after each run. Do not rely on run status alone.