# Vijil

> Source: https://agentsindex.ai/vijil

Enterprise AI agent trust platform for testing, guarding, and improving agents.

- Best for: Enterprise AI teams that need to prove agent safety before production and keep it enforced afterward.
- Look elsewhere if:
  - you want any plan without talking to sales
  - your data must stay in your region below the In your network plan

Starting price

Free

From [vijil.ai/pricing](https://www.vijil.ai/pricing/), checked 16 November 2026

- MCP server Yes From docs.vijil.ai/developer-guide/agentic/tools, 16 November 2026 [docs.vijil.ai](https://docs.vijil.ai/developer-guide/agentic/tools)
- API Yes From docs.vijil.ai/developer-guide/agentic/quickstart, 16 November 2026 [docs.vijil.ai](https://docs.vijil.ai/developer-guide/agentic/quickstart)
- Open source Not stated
- Self-host Yes From vijil.ai/pricing, 16 November 2026 [vijil.ai](https://www.vijil.ai/pricing/)

[Docs](https://docs.vijil.ai/)

What we checked

- Pages last read: 16 November 2026
- Facts checked: 18 facts · From 14 sources
- AI answers: Too few read · 58 buyer-question answers read, November 2026

[How we check](https://agentsindex.ai/methodology) [Editorial policy](https://agentsindex.ai/editorial-policy#not-for-sale)

[Visit Vijil](https://agentsindex.ai/go/vijil) [Compare](https://agentsindex.ai/compare?cat=security-ai-agents&products=vijil)

## At a glance

- Category: [Securing AI use](https://agentsindex.ai/categories/security-ai-agents)
- Tags: Private AI, Self-hosted, SOC 2

### For developers

- Works with: LangChain, CrewAI, Google ADK, Agent Development Kit, Amazon Bedrock, Microsoft Copilot and 10 more

## Vijil in our research

[All research](https://agentsindex.ai/blog)

Study 19 October 2026 – 16 November 2026

### [AI Tool Pricing Index, November 2026](https://agentsindex.ai/data/ai-tool-pricing-index)

Pricing

Contact sales

82 of the 361 tools we checked ask you to contact sales for a price.

Edition price, checked 13 November 2026.

Pricing pages of the AI tools, November 2026 edition as of 16 November 2026 n = 361 tools checked

## What is Vijil?

Vijil is an enterprise AI agent trust platform for teams that need to evaluate, protect, and continuously improve agents across the full lifecycle. It combines VIJIL DEPOT, VIJIL DIAMOND, VIJIL DOME, and VIJIL DARWIN to test workflows, enforce policy in real time, and learn from production telemetry. It supports on-premises deployment, multi-language SDKs, and a REST API, and is used by SmartRecruiters, Google, Autonoma, and Digital Ocean.

## Four modules around one trust loop, and one is not shipped

Vijil sells a loop with four parts, each named on [its home page](https://www.vijil.ai/): Discover "gives every agent an identity it can prove", Diamond "verifies it with tests on which it never trained", Dome "holds the limits it cannot cross", and Darwin "now in development, will evolve it inside them".

**Discover** scans code, cloud accounts and endpoints for agents, including unsanctioned ones, and gives each one an identity based on the CNCF SPIFFE standard.

**Diamond** compiles a test suite from an agent's purpose, personas and policies, runs it against the live agent and returns a Trust Score. On [the Diamond page](https://www.vijil.ai/product/diamond/) that score is computed on "the worst-scoring 10% of responses", not the average. The same page reports a model whose mean of 92 hid a Trust Score of 48.3. Adaptive Red Teaming, where attacker agents run successive waves against the live agent, is marked "launched October 7, 2026".

**Dome** is the runtime layer. It guards inputs, LLM calls, tool calls and outputs, and it mounts on a Python function with a decorator. It is open source under Apache-2.0.

**Darwin** reads an agent's code and telemetry and proposes fixes as pull requests. [Its own page](https://www.vijil.ai/product/darwin/) says "Darwin is not yet available to customers: today it runs only against repositories we configure."

The Dome benchmark on the home page covers eleven models, measured in September 2026. It reports a median gain of 17.6 points on their worst responses, and one model, Kimi K3, did not move. Vijil ran those measurements on its own taxonomy, and no outside party reproduced them.

## One agent is free, and no paid tier has a published figure

[The pricing page](https://www.vijil.ai/pricing/) has four columns. Open source costs "Nothing, forever" and is bounded only by your hardware. Hosted Free covers "One agent; a baseline evaluation a week, a Quick adaptive run a month" on console.vijil.ai. Hosted Team is listed as "Per agent", billed monthly or annually. In your network is "Annual, quoted once". The two paid columns end in a "Talk to us" button, and no dollar amount appears anywhere on the page.

The metering rules are stated plainly: "never per evaluation or per token" on Team, and "No count, no token meter" in your own network. Discovery is free at any scale on every column. The free tier "holds recent history rather than a year of it, and it will not export." Evaluating on every change, more agents, and the Standard and Deep adaptive efforts all require Team.

The [services page](https://www.vijil.ai/services/) sells a separate six-week guided pilot on one agent at a "Fixed price, quoted per pilot, before we start". After the pilot come Base and Scale engagements, also with no price. Only the Scale engagement lists "SSO, SLAs and a named technical account manager". The product pages never mention SSO or an uptime commitment.

## Vijil's own pages disagree on signup, export and per-agent pricing

[The free signup page](https://www.vijil.ai/free-trial/) says "Create a free account... No sales call, no credit card." Today you ask and we open an account, usually the same day." The home page sits between the two: "The hosted console takes a short form and opens the same day."

Read together, a report download seems to be allowed while the evaluation history stays on the console, but neither page says that.

The [terms of service](https://www.vijil.ai/policies/terms-of-service/), last modified October 5, 2026, grant free use "solely for your internal evaluation purposes and not for production or commercial use" and reserve the right to "turn off your access to the Free Trial or Free Tier at any time".

The services page promises "Flat price for all your agents. No per-seat traps, no token-based blowouts, no per-agent novelties." The pricing page prices Team "Per agent". The services page frames its tiers as engagements priced apart from the product, but the two sentences still read against each other.

## Hosted data sits in the US; the annual tier keeps it in your network

The [Trust Center](https://www.vijil.ai/policies/trust-center/) says that data is "hosted on Google Cloud Platform (GCP) databases. These databases are all located in the US." No region choice is offered. The in-network tier runs in a VPC, a datacenter or an air-gapped network, and Vijil says it is "designed to run in a network with no egress at all".

The Trust Center names a SOC 2 Type II attestation, at least annual third-party penetration testing, encryption at rest and TLS 1.2 in transit. The Diamond reports are pitched at auditors working under the EU AI Act, NIST AI RMF and ISO 42001.

The comparison table on the same page adds that the Discover endpoint installers are open source, while the Diamond client is "Free, not open source."

## Founded in 2023 by former AWS staff, with $23M raised

Vijil, Inc. is based at 700 El Camino Real, Menlo Park, California, according to its terms. [The company page](https://www.vijil.ai/company/) gives 2023 as the founding year. It names Vin Sharma as founder and CEO, previously GM and Director of Engineering for AWS SageMaker, and Tim G. J. Rudner, an assistant professor at the University of Toronto, as chief scientist. Its advisors include Leon Derczynski, founder of garak and NeMo Guardrails, and Ruslan Salakhutdinov of CMU.

A [November 2025 announcement](https://www.vijil.ai/blog/vijil-raises-17-million-to-make-ai-agents-resilient-named-a-gartner-cool-vendor/) reports $17 million led by Brightmind Partners, with Mayfield and Gradient participating, for $23 million in total.

## Customer evidence comes from Vijil's own case studies

[The case studies page](https://www.vijil.ai/case-studies/) names five: SmartRecruiters' Winston hiring agent, NEAR AI certifying third-party agents before they transact, a Brainforest property search, an Autonoma IoT troubleshooting agent, and DigitalOcean's GenAI Platform. SmartRecruiters' SVP of engineering is quoted saying Vijil helps it "ship AI agents in six weeks instead of six months". Each study is published by Vijil, and the page lists no independent review.

## Where it fits, and where it does not

Vijil fits a team running several agents that needs a test score and a runtime control a risk owner will sign off on. It also fits a team that cannot let agent traffic leave its own network. The Python decorator, CLI and Claude Code plugin all assume a developer is involved. A buyer who needs SSO, an uptime commitment or EU data hosting on the hosted console will not find any of them on the product pages, and the self-improving Darwin module is not yet sold.

Our take

Check whether your deployment needs on-premises handling of prompts and data before you commit. Vijil says it supports on-premises deployment, so confirm that requirement is part of your intended setup and not just an optional preference.

## Vijil price and terms

### Terms

- Free plan: Yes
- Card to start: Not required
- Sign up without sales: Some plans

The terms in full

One agent on the hosted console with no expiry: one baseline evaluation a week and one Quick adaptive run a month, recent history only, no export; accounts opened by Vijil by hand. Terms limit free use to internal evaluation, not production or commercial use. Dome (Apache-2.0) is free to run in your own process with no account.

### For AI agents

- MCP server: Yes [docs.vijil.ai, 16 November 2026](https://docs.vijil.ai/developer-guide/agentic/tools)
- API: Yes [docs.vijil.ai, 16 November 2026](https://docs.vijil.ai/developer-guide/agentic/quickstart)

[Developer docs](https://docs.vijil.ai/)

### What needs a higher plan

- In your network: Retention you control, Data stays in your region

[![Vijil](https://agentsindex.ai/img/1024/screenshots/vijil/pricing.webp)](https://agentsindex.ai/img/1920/screenshots/vijil/pricing.webp)

Pricing page, captured 30 October 2026

[vijil.ai/pricing](https://www.vijil.ai/pricing/), checked 16 November 2026

### All plans and limits

| Plan | Price | Main limits |
| --- | --- | --- |
| Hosted — Free | $0 a month | Agents: One agent; Baseline evaluations: a baseline evaluation a week; Discover: Any scale |
| Open source | $0 a month | Bounded by: Your hardware |
| Hosted — Team | Price on request | The agents you choose |
| In your network | Price on request | Agents and evaluations: Every agent, every evaluation. No count, no token meter. |

Read 9 November 2026 from [vijil.ai/pricing](https://www.vijil.ai/pricing/).

## What we checked on Vijil

Yes Partly No

What it costs to start Yes

### What the price counts

Hosted Team tier is priced per agent, never per evaluation or token; the in-network deployment is a flat annual subscription with no agent count. No figures are published for either.

> On Team, a price per agent, monthly or annual, with no per-evaluation meter and no token bill. In your own network you buy a deployment rather than a…

FAQ, 'So what does it actually cost?' [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

What it costs to start Yes

### Free to use

One agent on the hosted console is free with no expiry, and the Dome engine is free open source; the free hosted tier is capped and its terms limit it to internal evaluation, not production use.

> For one agent on our console, nothing.

FAQ, 'So what does it actually cost?' [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

What it costs to start Partly

### Self-serve signup

The open-source Dome engine installs from PyPI with no account, but the hosted free tier still goes through Vijil, which opens each account by hand, usually the same day; Team and in-network tiers start with a sales conversation.

> Is the free tier self-serve? Not yet. Today you ask and we open an account, usually the same day.

FAQ, 'Is the free tier self-serve?' [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

What it costs to start No

### Card to start

Starting the free hosted account or the open-source engine asks for no credit card, and the vendor says so on both its home page and its free signup page.

> Create a free account, point Vijil at your agent, and get a complete trust report the same day. No sales call, no credit card.

Hero paragraph under 'Test your agent before you trust your agent' [vijil.ai/free-trial](https://www.vijil.ai/free-trial/) Checked 16 November 2026

What it costs to start Partly

### Contract

Hosted Team can be billed monthly or annually, while running the platform in your own network is sold only as an annual subscription quoted once; no seat minimum is stated.

> In your own network you buy a deployment rather than a quantity: an annual subscription for the platform, with no agent count.

FAQ, 'So what does it actually cost?' [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

What it works with Yes

### Works with

Dome ships adapters for LangGraph, Google ADK, Strands and MCP, plus a Claude Code plugin; integration guides cover LangChain, CrewAI and Google ADK, and the SDK wraps any Python agent. Available on every tier including open source.

> It ships adapters for LangGraph, ADK, Strands and MCP, and Claude Code users can add the Vijil plugin from the editor.

Section 'Install it, wrap a step, watch what it stops' [vijil.ai/product/dome](https://www.vijil.ai/product/dome/) Checked 16 November 2026

What it works with Yes

### Public API

A documented REST API with bearer-token auth sits beside a Python SDK and CLI, covering agent registration, evaluations, red team campaigns and Dome configs in a public API reference.

> REST API Custom integrations and non-Python applications An HTTP client and a bearer token

Quickstart, interface table, REST API row [docs.vijil.ai/developer-guide/agentic/q…](https://docs.vijil.ai/developer-guide/agentic/quickstart) Checked 16 November 2026

What it works with Yes

### MCP server

Vijil publishes its own MCP server, the vijil-mcp package, with documented tools for agents, evaluations and teams; it is documented for Claude Code and installed locally rather than as a hosted URL.

> Complete reference for all tools exposed by the Vijil MCP server, organized by workflow.

Use MCP Tools page, opening description [docs.vijil.ai/developer-guide/agentic/t…](https://docs.vijil.ai/developer-guide/agentic/tools) Checked 16 November 2026

What it does on its own Yes

### Acts on its own

The discover, verify and defend loop runs automatically across an estate, with Dome allowing, steering or denying each agent step at runtime; Darwin, which proposes agent rewrites as pull requests for approval, is still in development.

> So the work of trusting it never finishes. Vijil runs that loop automatically.

Section 'The Trust Loop' [vijil.ai](https://www.vijil.ai/) Checked 16 November 2026

What it does on its own Yes

### Limits on what it may do

Policy written in YAML compiles into Controls that allow, steer or deny each model or tool call before it runs, with a shadow mode that logs denials without blocking, and per-agent tool allow lists where SPIFFE runs.

> Each Control allows, steers or denies a step, before or after it runs. Shadow mode logs what it would deny without blocking anything.

Section 'One policy, three kinds of control', step 1 'Policy becomes Controls' [vijil.ai/product/dome](https://www.vijil.ai/product/dome/) Checked 16 November 2026

What it does on its own Partly

### No code needed

Agent owners can register agents, run evaluations and configure Dome from the web console, but enforcing Dome inside an agent's process means adding the Python package to its code.

> You use the Vijil console to register Agents, run Evaluations, configure Dome, monitor Dome Metrics, and report to stakeholders.

Continue with Vijil, 'Agent Owner's Guide' card [docs.vijil.ai/concepts/welcome](https://docs.vijil.ai/concepts/welcome) Checked 16 November 2026

What happens to your data Yes

### No training on your data

The vendor states it does not train on customer data on any tier, free or paid, and offers to help customers train on their own data inside their own network.

> We do not train on customer data — not on the free tier and not on any other.

FAQ, 'Do we have to send you our data?' [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

What happens to your data Partly

### Retention you control

Hosted data is kept as long as the vendor deems necessary and the free tier holds only recent history; full control over retention comes from the in-network tier, where decision traces stay on your own hardware.

> Every agent, with decision traces kept in your network. No per-agent count, no token meter

Comparison table, Dome row, 'In your network' column [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

What happens to your data Partly

### Data stays in your region

Hosted tiers store data on Google Cloud databases located in the US with no region choice; keeping data elsewhere means the annual in-network tier in your own VPC, datacenter or air-gapped network.

> All of our data is hosted on Google Cloud Platform (GCP) databases. These databases are all located in the US.

Cloud Security, 'Data Hosting Security' [vijil.ai/policies/trust-center](https://www.vijil.ai/policies/trust-center/) Checked 16 November 2026

What happens to your data Yes

### Self-hosted

Dome runs free in your own process, and the full platform runs in your VPC, on premises or air-gapped under the annual in-network subscription, with a documented Kubernetes deployment guide.

> In your own network, no: the platform runs inside your VPC or on your own hardware, and it is designed to run in a network with no egress at all.

FAQ, 'Do we have to send you our data?' [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

What happens to your data No public code found

### Open source

Its GitHub repository holds issues or documents, no code

> Make the agents you build more reliable, secure, and safe — from inside Claude Code. Inspect (free, no account), evaluate, analyze, protect, adapt,…

[github.com/vijilAI/vijil](https://github.com/vijilAI/vijil) Checked by us, 15 November 2026

What happens to your data Partly

### Take your data out

The free hosted tier explicitly will not export history; export on Team is not described, though trust reports can be downloaded and a REST API reads evaluation results.

> It holds recent history rather than a year of it, and it will not export. Nothing about it expires.

FAQ, 'What does the free tier not do?' [vijil.ai/pricing](https://www.vijil.ai/pricing/) Checked 16 November 2026

Whether a company can buy it Yes

### Security audits

Vijil names a SOC 2 Type II attestation and annual third-party penetration testing on its Trust Center; no ISO 27001, ISO 42001 or published DPA is named.

> Vijil holds a SOC 2 Type II attestation.

Organizational Security, 'Information Security Program' [vijil.ai/policies/trust-center](https://www.vijil.ai/policies/trust-center/) Checked 16 November 2026

Whether a company can buy it Yes

### Audit log

Trust Runtime emits structured audit events for guard decisions, tool access checks, attestation results and enforcement downgrades, routable to your own telemetry; a log of console user actions is not described.

> Trust Runtime emits an AuditEvent for security-relevant decisions. By default, it sends events to Python logging.

Section 'Capture Audit Events' [docs.vijil.ai/developer-guide/protect/t…](https://docs.vijil.ai/developer-guide/protect/trust-runtime) Checked 16 November 2026

## Questions about Vijil

- How much does Vijil cost?: Vijil has a free plan. Its paid plans are Hosted — Team and In your network. From vijil.ai/pricing, checked 16 November 2026.
- Does Vijil have an API or an MCP server?: Vijil has an API. From docs.vijil.ai/developer-guide/agentic/quickstart, checked 16 November 2026. Vijil has an MCP server. From docs.vijil.ai/developer-guide/agentic/tools, checked 16 November 2026.
- Is Vijil open source, and can you host it yourself?: You can host Vijil yourself. From vijil.ai/pricing, checked 16 November 2026.
- What have we checked on Vijil?: 18 facts, under these headings: What it costs to start, what it works with, what it does on its own, what happens to your data, whether a company can buy it.
- How do we check Vijil?: We hold 18 facts on Vijil, quoted from 14 pages. Facts and price last checked 16 November 2026, when we also found the last change.

[How we check every listing](https://agentsindex.ai/methodology)

## Sources and changes

Last changed 16 November 2026

Sources 14

- Read 16 November 2026 · [vijil.ai](https://www.vijil.ai/pricing/) /pricing
- Read 9 November 2026 · [vijil.ai](https://www.vijil.ai/product/discover/) /product/discover
- Read 16 November 2026 · [vijil.ai](https://www.vijil.ai/product/dome/) /product/dome
- Read 16 November 2026 · [vijil.ai](https://www.vijil.ai/policies/trust-center/) /policies/trust-center
- Read 9 November 2026 · [vijil.ai](https://www.vijil.ai/product/diamond/adaptive_redteam/) /product/diamond/adaptive_redteam
- Read 8 November 2026 · [docs.vijil.ai](https://docs.vijil.ai/) /
- Read 8 November 2026 · [vijil.ai](https://vijil.ai/) /
- Read 16 November 2026 · [vijil.ai](https://www.vijil.ai/) /
- Read 16 November 2026 · [docs.vijil.ai](https://docs.vijil.ai/concepts/welcome) /concepts/welcome
- Read 16 November 2026 · [docs.vijil.ai](https://docs.vijil.ai/developer-guide/agentic/quickstart) /developer-guide/agentic/quickstart
- Read 16 November 2026 · [docs.vijil.ai](https://docs.vijil.ai/developer-guide/agentic/tools) /developer-guide/agentic/tools
- Read 16 November 2026 · [docs.vijil.ai](https://docs.vijil.ai/developer-guide/protect/trust-runtime) /developer-guide/protect/trust-runtime
- Read 16 November 2026 · [vijil.ai](https://www.vijil.ai/free-trial/) /free-trial
- Read 9 November 2026 · [vijil.ai](https://www.vijil.ai/product/darwin/) /product/darwin

What changed 1

- 16 November 2026 What it does

## Vijil price history

Vijil was free on both of our reads, from 30 October 2026 to 16 November 2026.

| Price | Plan | First seen | Last seen | Readings | Source |
| --- | --- | --- | --- | --- | --- |
| Free |   | 30 October 2026 | 16 November 2026 | 2 | [Pricing page, read 30 October 2026](https://www.vijil.ai/pricing/) |

From our research, read twice since 30 October 2026. A date is when we read the page, not when the price changed.

## Similar to Vijil

- [Base2026](https://agentsindex.ai/base2026) Free SEO research MCP server, public API and browser SEO checks with original sources attached. Free plan
- [CurlShot](https://agentsindex.ai/curlshot) Screenshot and PDF rendering API with an MCP server, so an agent can see any web page. Free plan, then $19 a month
- [Bot Street](https://agentsindex.ai/botstreet) Marketplace where AI agents post offers, take paid tasks and run shops, reached over MCP, a CLI or REST. Free plan

## Is Vijil yours?

Free, and the page then carries the verified owner mark. You confirm it with an email at vijil.ai. [Claim this listing](https://agentsindex.ai/claim/vijil?from=owner-line)

[Report an error](mailto:hello@agentsindex.ai?subject=Report%20an%20error%3A%20Vijil%20(vijil)&body=What%20is%20wrong%20on%20the%20page%2C%20and%20where%20on%20the%20product) if a fact is wrong.
