SkillAgentSearch skills...

olivares

Ground truth for enterprise AI: your AI coding tools, your accounts and every session on your own server, from one web console, CLI or API. Self-hosted, open source.

Install / Use

claude mcp add olivaresai -- npx -y github:olivaresai/olivares

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

84/100

Category

Security

Supported Platforms

Claude Code
Claude Desktop

Our assessment of olivares

olivares scores 84/100 on our quality scale, 844th of 1,103 Security skills we index.

Its MCP Server is 13 KB long, well organised into 9 sections with 9 code examples: a thorough specification that gives an agent plenty to work with.

It has 10 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
30/30
Structure
20/20
Description
15/15
Adoption
4/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated today, so olivares is actively maintained.
  • It is released under AGPL-3.0, a copyleft license: you can use it, but modified versions you distribute must carry the same license.
  • Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

olivares compared with similar skills

All 4 of these similar skills score higher than olivares; compare them before choosing.

SkillScoreStarsUpdatedFormat
olivares (this skill)by olivaresai8410todayMCP Server
Agent-Reachby Panniantong10097.0k5d agoCLAUDE.md
headroomby headroomlabs-ai10075.2ktodayCLAUDE.md
CowAgentby zhayujie10047.4ktodayCLAUDE.md
Scraplingby D4Vinci10087.3ktodayMCP Server

Frequently asked questions

How do I install olivares?
Run claude mcp add olivaresai -- npx -y github:olivaresai/olivares. The install tabs above show the steps for each supported agent.
Which AI agents does olivares work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is olivares safe to use?
It is AGPL-3.0-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is olivares still maintained?
The repository was last updated today, so olivares is actively maintained.
<div align="center">

<a href="https://olivares.ai"><img src=".github/assets/olivares-banner.png" alt="Olivares AI — Ground truth for enterprise AI" width="720"></a>

Languages: English · Español · 简体中文 · Русский · 日本語 · Deutsch · Français

Run the AI your team already uses, with the same control you have over the rest of your infrastructure.

What it does · Install · Editions · Documentation · Community · olivares.ai

License: AGPL-3.0-only SDK & connectors: Apache-2.0 <!-- release --> Next release: 0.1<!-- /release --> Status: beta Contributor Covenant

</div>

Your developers work with Claude Code and Codex. Agents call MCP servers, models and internal APIs, and scheduled jobs run on their own. Each piece keeps its own logs and its own permissions, so simple questions have no quick answer: which agent changed this file, who approved it, what did AI cost us this month?

Olivares AI answers them in one place. It connects to the agents and tools you already use, shows what each one does, applies your rules before an action runs, and keeps a signed record of everything. It is a single program that runs on your own servers, and the complete product is free and open source.

<div align="center"> <img src=".github/assets/motion-access-map.gif" width="840" alt="Motion diagram of the read/write access map: agents, sessions and identities on the left, the resources they reach on the right, reads in blue, writes in orange, one observed write that was never permitted flagged as a drift finding."> <br><sub><b>The access map</b> — what every agent reads and writes, and the one write nobody allowed.</sub> </div>

What it does

  • Know what is running. Every agent, session, model, MCP server and tool in one inventory. The access map shows what each one reads and writes, and flags access that no rule allows.
  • Stop an action before it does damage. Olivares AI has four deny-closed enforcement points that check each action before it runs: inside Claude Code, in the model proxy, at every MCP tool call and between agents. A risky action waits for a second person, a forbidden one does not run, and one switch stops every agent at once. When a check cannot decide, the action does not run.
  • Keep AI spending in check. Budgets per team, agent or model warn, slow down or stop spending before the invoice arrives.
  • Give agents your company's knowledge, safely. Connect SharePoint, Confluence, Google Drive, Notion, Salesforce, Snowflake, S3 and PostgreSQL. Each agent sees only what the person behind it is allowed to see.
  • Keep work going across sessions. Tasks, owners and decisions stay when a session ends. Start, join and stop Claude Code, Codex and Grok sessions from the browser, without SSH.
  • Show proof when someone asks. Every decision goes into a signed, tamper-evident log. Business Compliance Packs map evidence to 26 framework catalogs and produce reports for your security team and auditors. Community retains stored evidence and its JSON/CSV exports.

It works with the tools you already have: Claude Code, Codex, Grok, Cursor, gemini-cli, opencode, OpenHands and local models through Ollama. 32 modules and 136 integrations: every module · every connector.

Community keeps local observability, stored settings and backup export. SIEM/ITSM push, external telemetry delivery and posture export are included in the Business base edition.

Session network isolation covers Claude Code and Codex sessions bound to a provider key. Such a session reaches only its provider's endpoint, through a per-session proxy in its own network namespace, or it does not start. Grok Build, OpenCode and account-login sessions can still reach any host the network permits. See session network isolation before running workloads that require exclusive access to one provider.

Install

Pick one method and copy its block. At the end, olivares quickstart prints the console address and a one-time token to create the first administrator. Every release is signed, and each method verifies what it downloads before it installs it (verify a download yourself).

Linux and macOS, one command. Detects your system, verifies the release, installs only the binary and never uses sudo.

curl -fsSL https://olivares.ai/olivares/install.sh | sh
olivares quickstart

Docker. Multi-arch. Container images are based on Debian 13 slim (with Node.js 24 for the agent tools) and run as a non-root user. It listens on every host interface; put 127.0.0.1: before each -p to keep it local.

<!-- release -->
docker run -d --name olivares -p 8443:8443 -p 8444:8444 \
  -v olivares-data:/var/lib/olivares \
  docker.io/olivaresai/olivares:0.1 \
  serve --listen :8443 --grpc-listen :8444 --data-dir /var/lib/olivares
<!-- /release -->

Docker Compose. SQLite on one node, with optional Postgres and backups.

git clone --depth 1 https://github.com/olivaresai/olivares.git && cd olivares
docker compose -f deploy/compose/docker-compose.yml up --wait --wait-timeout 120

Sessions work on one host folder: set OLIVARES_PROJECT_DIR to its absolute path before up, and Compose mounts it at /project. On hosts whose AppArmor policy limits user namespaces (Ubuntu 24.04 and later), load the session profile first: see Docker Compose.

Debian and Ubuntu. The package adds a no-login olivares user and a hardened service; you start it.

curl -fsSLO https://github.com/olivaresai/olivares/releases/download/0.1/olivares_0.1_linux_amd64.deb
sudo dpkg -i olivares_0.1_linux_amd64.deb && sudo systemctl enable --now olivares

RHEL, Fedora and SUSE.

curl -fsSLO https://github.com/olivaresai/olivares/releases/download/0.1/olivares_0.1_linux_amd64.rpm
sudo rpm -i olivares_0.1_linux_amd64.rpm && sudo systemctl enable --now olivares

Alpine.

curl -fsSLO https://github.com/olivaresai/olivares/releases/download/0.1/olivares_0.1_linux_amd64.apk
sudo apk add --allow-untrusted olivares_0.1_linux_amd64.apk && sudo rc-service olivares start

On ARM servers, use arm64 instead of amd64. Every file of the release: the release page.

Homebrew. macOS and Linux.

brew install olivaresai/tap/olivares && olivares quickstart

From source. Go 1.26+, Task and pnpm.

git clone --depth 1 https://github.com/olivaresai/olivares.git && cd olivares
task build && ./bin/olivares quickstart

Offline networks: bundle the signed image, chart and verification material, then install air-gapped. Windows has no native build yet: use the Docker image or WSL2. Upgrades and rollback: how-to. Every option in detail: INSTALL.md.

Try it first with demo data, on your machine only (the demo password is public):

olivares serve --seed-demo --insecure --listen 127.0.0.1:8901 --grpc-listen 127.0.0.1:8902 --data-dir "$(mktemp -d)"

Then open http://127.0.0.1:8901.

Helm, the Kubernetes operator, Terraform, appliance and FIPS/STIG deployment artifacts are distributed with Business; air-gapped installation requires Enterprise. See editions.

Editions and pricing

Community is the complete product, free and open source. Business adds what a company needs to run it in production. Enterprise is for groups with larger or regulated estates.

| | Community | Business | Enterprise | |---|---|---|---| | Price | Free, AGPL-3.0 | USD 129/month or USD 1,290/year | Annual contract | | What you get | The complete product: unlimited users and all four deny-closed enforcement points | Everything in Community, plus Regulated Operations, AI Runtime Security, Compliance Packs and Identity & Scale, the commercial licence, signed updates and email support | Everything in Business, plus more companies, deployments and identity providers, offline mirrors and support terms agreed with you | | Scope | One active identity provider | One company, one active instance at a time | Agreed in the contract |

Regulated Operations adds regulatory retention floors, legal-hold reconciliation on archives and WORM archives on Azure and GCS. AI Runtime Security adds deeper inspection of what agents send, receive and run. Compliance Packs provide framework assessments, regulatory views, OSCAL export and on-demand HTML/PDF reports, and draft the DORA register of information and the ISO/IEC 42001 pack for your auditor. Identity & Scale connects several identity providers at once and grows with larger deployments.

Compare editions and buy · What each edition includes · What is open and what is commercial

Architecture

One Go binary with the console built in. It serves a REST API, a gRPC API, the olivares command line. The Terraform provider is distributed with Business. Collectors run inside your network, and data stays in SQLite or PostgreSQL on your servers. How it fits together.

Documentation

docs.olivares.ai has install guides, a guide for every connector, recipes for common policies and the API reference. Start with What is Olivares AI. What runs today and what is still planned: Honesty & limits. Releases: GitHub · CHANGELOG.md.

On the website: product · solutions · how it works · architecture · security · trust · compare · demo · changelog · status · roadmap · brand · press.

Security

Found a vulnerability? Report it privately through SECURITY.md. Olivares AI records which agent touched which resource, not the content, and opening that record is itself logged. Licences are checked offline; the open-source core never calls us.

Community

Contributions are welcome. CONTRIBUTING.md explains the setup, the sign-off and how connectors fit in. Code of conduct · Support · Governance · Changelog.

Support the project

Olivares AI is built in the open. If it helps you, sponsor its development on GitHub Sponsors — olivaresai or fran-olivares — or buy us a coffee on Ko-fi. Sponsors who want to be named appear in SUPPORTERS.md. Sponsorship is not a support c

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars10
CategorySecurity
Updated2h ago
Forks1

Languages

Go

Trust signals

97/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 info
olivares — MCP Server: Install & Safety Check | SkillAgent