tools
Vectara Platform
Enterprise platform for building, deploying, and governing AI agents over business data.
Vectara helps organizations build AI agents that retrieve information from documents, generate answers, use tools, and apply governance controls. It supports multimodal indexing, neural and lexical search, reranking, customized generation, factual-consistency scoring, and tool validation.
Teams can deploy it as SaaS, in a customer-managed VPC, or on-premises. It provides APIs and integrations for data ingestion and agent development. Pricing is enterprise-focused: paid plans start at $100K/year, while the free trial lasts 30 days and is not a free-forever plan.
Features
- Build multiple AI agents on shared data infrastructure
- Index and retrieve text, tables, images, and complex documents
- Combine neural search, lexical search, reranking, and generation
- Add custom agent instructions and bring your own tools
- Use bring-your-own embedding and generative models
- Score factual consistency and validate agent tool use
- Deploy through SaaS, VPC, or on-premises environments
- Access platform capabilities through APIs
Use cases
- Build enterprise knowledge-base assistants
- Automate customer-support conversations and escalation
- Analyze technical documents for troubleshooting
- Manage document lifecycles with grounded retrieval
- Monitor agent usage, traces, costs, and performance
Pros
Cons
Latest updates
- Longer Session-Enrichment Tool Timeouts
Each session-enrichment tool call now accepts a timeout_seconds value up to 600 seconds (10 minutes), increased from 120 seconds.
- Skill Instructions Are Omitted from Tool Responses
The invoke_skill tool response now contains the skill name in skill_name without the former content property.
- Agent Evaluation Correctness Fixes
Agent evaluations no longer include responses generated before the user input being evaluated.
- Inline Images in Chat Completions
createChatCompletion accepts a base64 data URL in image_url.url and sends the decoded bytes to the model.
- Unreadable Page Keys Are Rejected as Bad Input
Malformed or unreadable page_key values are rejected with HTTP 400 and the field error invalid page_key at query.page_key.
Capabilities
- Choice of models — “Bring your own LLM, or use ours” source
- Self-hosted — “Vectara is the only agentic AI platform you can deploy as SaaS, in your VPC, or on-premises, with the same enterprise-grade security everywhere.” source
- API — “REST API Apps and services” source
- Runs models for you — “Vectara domain LLM Built-in” source
- Builds agents and workflows — “Agent harness: skills, tools, MCP, memory” source
- Search over your data — “Semantic index (vector search)” source
- Traces and evaluates — “Full observability with verifiable data trace in one click, from answer to source” source
Get it
Pricing
- Starting price
- Free
- Prices checked
- 2026-11-06
- Read
- by web search
30 Day Free Trial
Free
- 10,000 credits included
- All features included for 30 days
SaaS
Price on request
- 1 SaaS Deployment
- 10,000,000 credits/year included with base platform package purchase
VPC
Price on request
- 1 VPC Deployment (Any VPC)
- 100,000,000 credits/year included with base platform package purchase
On-prem
Price on request
- 1 On-Premise Deployment
- 100,000,000 credits/year included with base platform package purchase