# Autonomous AI Agent

> Build your autonomous AI agent module by module and see the fixed price update: base from ₪25,000, typical configured builds ₪45,000-75,000, plus VAT as required by law. Save it as a formal itemized proposal.

- Web page: https://sporesec.com/en/packages/ai-agent
- Language: English
- Last updated: 2026-10-10
- Also in עברית: https://sporesec.com/he/packages/ai-agent.md

## Answer

Build an autonomous AI agent on SporeSec module by module and watch the fixed price update. The agent retrieves grounded knowledge, plans and reasons, acts through your tools, and stays inside guardrails with human handoff. The base starts from ₪25,000 and typical configured builds cost ₪45,000-75,000, plus VAT as required by law.

## What you get

- Retrieve grounded knowledge
- Plan and reason
- Act through your tools
- Guardrails and human handoff

## Pricing

- from ₪25,000
- typical build: ₪45,000-75,000
- ₪750/mo optional care
- plus VAT as required by law

## Modules and prices

### Knowledge and retrieval

What the agent knows, and how it stays current. The base grounds answers in a starter set; this is the real retrieval brain over your content.

- **Knowledge base**: How the agent knows your business beyond the starter materials in the base.
  - Starter materials (included): The materials you provide, indexed and grounded, already in the base; no added retrieval. (included)
  - Retrieval over your documents (RAG): A retrieval layer over a document library you provide (policies, guides, specs, past tickets). The build indexes and grounds answers; discovery covers curating and structuring the source set. (₪4,500 (₪3,000 + ₪1,500))
  - Retrieval over structured data: Structured retrieval over a product, price, or inventory catalog, kept in sync with your source of truth. (₪5,000 (₪3,000 + ₪2,000))
  - Hybrid knowledge base: A knowledge base spanning documents, structured data, and live systems together, with freshness rules for each source. Scope-dependent. (from ₪6,000)
  - Company brain: The company brain: your documents, catalogs and live systems indexed as one knowledge layer, refreshed as they change, with every answer fenced to those sources and a handoff when the brain does not know. Priced as a fixed set-up plus a monthly upkeep fee; the model usage bill from the LLM provider is separate. Upkeep: ₪550 a month. (₪5,000)
- **Automated knowledge refresh**: A re-indexing pipeline so the knowledge base stays current as your content changes, with no manual re-upload. (₪2,000)

### Tools and actions

What the agent can DO, not just answer. Each tool lets it act in a real system, with auth and safe-write guards.

- **Tool integration**: Each external system the agent can act in (create a record, book a slot, issue a document, trigger a workflow): a system we know, with a documented API or a ready connector. One system per unit. (₪2,500, x6 max)
- **Custom tool (no ready connector)**: A bespoke tool over a system with no ready connector: the agent calls your API. The build is contained; the discovery share covers learning your endpoints, auth, data model, and edge cases. (₪4,000 (₪2,500 + ₪1,500), x4 max)
- **High-stakes actions**: Actions with real consequences (a payment, a refund, a contract issued) wrapped in confirmation, idempotency, and an audit trail, so a re-run never double-acts. Scope-dependent. (from ₪3,500)

### Channels and surfaces

Where the agent lives. The base ships a web chat widget; add the channels your users and staff actually use, all sharing one brain.

- **Primary channel**: The main surface the agent is built around.
  - Web chat widget (included): An embeddable chat widget on your site, already in the base. (included)
  - WhatsApp (Meta Cloud API): The agent on WhatsApp via the official Meta Cloud API, with the consent, template, and quality-rating handling that channel requires. (₪3,000)
  - Slack or Microsoft Teams: An internal agent inside Slack or Teams for your staff, wired to your workspace. (₪2,500)
- **Additional channel**: Each additional channel the same agent serves (web, WhatsApp, Slack, email, SMS), sharing one brain and memory. (₪2,000, x4 max)
- **Voice channel**: An inbound or outbound voice channel (phone or in-app) with speech-to-text and text-to-speech, scoped to your use case. Scope-dependent. (from ₪5,000)

### Memory and context

How much the agent remembers. The base holds the current conversation; go further with long-term memory and per-user personalization.

- **Memory model**: How far the agent remembers beyond the current thread.
  - Conversation memory (included): Within-conversation memory: the agent holds the current thread, already in the base. (included)
  - Long-term memory: Memory across conversations: the agent recalls a returning user, prior requests, and outcomes, so nobody repeats themselves. (₪3,000)
  - Personalization: Long-term memory plus per-user personalization: tailored responses, segments, and recall wired to your user records. (₪4,500)

### Reasoning and orchestration

How the agent thinks. The base is a single capable agent with tool use; deepen it to multi-step planning or a coordinated multi-agent system.

- **Orchestration**: How complex a task the agent can carry on its own.
  - Single agent (included): One capable agent with tool use, already in the base. (included)
  - Multi-step planning: The agent decomposes a goal, runs a sequence of tool calls, and self-checks the result before answering. (₪4,000)
  - Multi-agent system: Specialized agents coordinated for complex workflows: a planner, workers, and a reviewer, each with its own tools. Scope-dependent. (from ₪7,000)

### Guardrails, safety and evaluation

How tightly the agent is bounded and how you know it stays good. The base carries baseline guardrails; these harden and measure them.

- **Guardrail tier**: How strictly the agent is bounded and what it does when unsure.
  - Standard (included): The base defenses: answers grounded to your knowledge, prompt-injection defenses, PII minimization, and refusal out of scope. (included)
  - Strict: An allowed-topics allowlist, a tightened policy against made-up answers, sensitive-data redaction, and refusal logging for regulated or sensitive contexts. (₪3,000)
- **Evaluation harness**: A graded test set of real prompts with regression scoring on every change and a quality dashboard, so the agent improves without silent regressions. The discovery share covers building the test set from your cases. (₪4,000 (₪2,500 + ₪1,500))

### Human handoff and oversight

How a person stays in control. The base pauses and alerts you when unsure; these add routing, a review console, and approval gates.

- **Handoff model**: What happens when a conversation needs a person.
  - Simple handoff (included): The agent pauses and alerts a person with full context, already in the base. (included)
  - Rules-based routing: Route escalations by topic, business hours, or agent confidence, each to the right person or team. (₪2,500)
  - Review console: A console where staff approve, edit, or take over agent responses, with assignment and status, so a team handles oversight together. Build plus discovery of your review workflow. (₪3,500 (₪2,500 + ₪1,000))
- **Human approval gate**: A human approval step before the agent takes a high-stakes action, with the decision recorded, so nothing irreversible happens unattended. (₪2,500)

### Analytics and observability

What you can see once the agent is live. The base logs every run; these turn logs into tracing, dashboards, and cost control.

- **Conversation tracing**: Full step-level tracing: every tool call, retrieval, and model decision captured, so you can see exactly why the agent did what it did and debug a bad answer. (₪3,000)
- **Analytics dashboard**: A dashboard and periodic report: volume, resolution rate, handoff rate, tool-success rate, and outcome tracking against your goals. (₪2,500)
- **Cost monitoring and budgets**: Per-conversation model-cost tracking with budget alerts, so token spend never surprises you and a runaway loop is caught early. (₪1,500)

### Deployment, hosting and scale

Where the agent runs and how it holds up under load. Managed cloud is the fast path; self-hosting or a private model keeps everything on your infrastructure.

- **Hosting**: Where the agent is hosted and which model it runs against.
  - Managed cloud (included): Deployed to managed cloud on your own domain, with nothing for you to operate, already in the base. (included)
  - Self-hosted (your cloud account): Deployed into your own cloud account, so the infrastructure and the model billing are yours. Build plus discovery of your environment and access. (₪4,500 (₪3,000 + ₪1,500))
  - Private or self-hosted model: Runs against a private or self-hosted model (open-weights or your own tenant) for data that cannot leave your boundary. Scope-dependent. (from ₪6,000)
- **Scale and load hardening**: Load and concurrency hardening for high traffic: caching, rate management, and a cost-at-volume review, so the agent stays fast and affordable as usage grows. Scope-dependent. (from ₪4,000)

### Care, continuity and extras

What happens after launch: getting your team running it, extending the warranty, a service commitment, and anything not on this sheet.

- **Team training and handover**: A hands-on training session for your team plus written runbooks, so the people who operate the agent day to day actually know how. (₪2,000)
- **Extended warranty**: The 30-day warranty extended to 90 days: fixes for anything that does not work as specified stay on us for longer after launch. (₪2,500)
- **Service-level agreement (SLA)**: A named response-and-resolution commitment with defined support hours, built on the tracing that detects issues. Scope-dependent. (from ₪3,000)
- **Something not on this sheet**: A capability you need that is not listed here. It rides the proposal as an open item, and we scope and price it on the call. (priced on a call)


## Facts

- The autonomous AI agent starts from ₪25,000, fixed price per scope, plus VAT as required by law. (source: registry:CATALOGS.aiagent.baseIlsFrom; as of 2026-10-10)
- Typical configured autonomous agent builds cost ₪45,000-75,000, plus VAT as required by law. (source: registry:CATALOGS.aiagent.typicalBuildIls; as of 2026-10-10)

## Next step

- Configure and price it online: https://sporesec.com/en/packages/ai-agent
- Book a free intro call: https://sporesec.com/en/book.md

---

- Contact: contact@sporesec.com | WhatsApp https://wa.me/972532322318 | https://sporesec.com/en/contact.md
- Book a free intro call: https://sporesec.com/en/book.md
- Machine-readable site data: https://sporesec.com/llms.txt | https://sporesec.com/agents/en/index.json