Raven MCP server
Commit Nov 13, 2026Connects AI assistants to enterprise DevOps tools like Jira, Bitbucket, and Azure DevOps with built-in PII scrubbing and autonomous error triaging.
Also known as bcgov/raven
- Stars
- 8
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 13, 2026
- License
- Apache-2.0
What is Raven MCP?
Raven connects AI assistants to enterprise DevOps tools like Jira, Bitbucket, and Azure DevOps with built-in PII scrubbing and autonomous error triaging.
Summary generated from the project's README
About Raven MCP server
Connecting local AI assistants to enterprise DevOps ecosystems is what Raven enables for developers in the BC Gov NR Sector. It provides a local-first analysis suite that bridges LLMs with tools like Jira, Confluence, Bitbucket, Azure DevOps, and Jenkins. By running locally via MCP servers, it allows users to manage the entire software development lifecycle, from searching JQL in Jira and navigating Confluence space trees to reviewing pull requests in Bitbucket or inspecting Jenkins build queues.
Security is a core focus, as Raven-owned servers utilize PiScrubber to remove personal information by default. While Atlassian tools share a single authentication session via SiteMinder SSO, other services like SharePoint, GitHub, and Artifactory require individual credentials stored in a local environment file. The suite also includes an autonomous DevOps pipeline that can detect production errors via SSH, triage them using AI, and automatically generate code fixes and pull requests.
What you can do with it
- 1Automated error triagingAn AI client can detect production errors, analyze them, and create pull requests with code fixes automatically.
- 2End-to-end pull request managementDevelopers can use the Bitbucket integration to review diffs, comment on code, and merge or decline pull requests.
- 3Centralized project documentation retrievalUsers can search and read project requirements and design documents stored in SharePoint Online.
- 4Infrastructure and asset queryingTeams can query the Jira Assets CMDB to inspect applications, environments, and technical stacks.
Written from the project's README. Every use case is checked against a quote from it.
Configuration
ADO_BASE_URLADO_DEFAULT_COLLECTIONADO_DEFAULT_PROJECTADO_PATARTIFACTORY_EMAILARTIFACTORY_PASSWORDARTIFACTORY_URLATLASSIAN_BASE_URLATLASSIAN_EMAILATLASSIAN_PASSWORDGITHUB_REPOSITORY_ALLOWLISTGITHUB_TOKENIMIS_CSV_PATHJARVIS_BASE_URLJARVIS_TOKENJENKINS_URLJIRA_EPIC_LINK_FIELDOLLAMA_FLASH_ATTENTIONOLLAMA_KV_CACHE_TYPERAVEN_AUTH_DEBUGRAVEN_SCRUB_PIRFCBUDDY_PATRFCBUDDY_URLSERVER_A_PASSWORDSHAREPOINT_URLSONARQUBE_TOKENSONARQUBE_URLSONAR_SCANNER_BINSSH_KEY_HOSTSSSH_KEY_PASSPHRASESSH_KEY_PATH
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablebcgov/raven is public and reachablegithub.com/bcgov/raven ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-13github.com/bcgov/raven ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubRAVEN - Resource Analytics, Visibility & Enterprise Navigator
Local-first AI analysis suite for BC Gov NR Sector. Connects Jira, Confluence, and Bitbucket behind SiteMinder SSO to an AI client via locally run MCP servers. Raven-owned servers that use PiScrubber scrub personal information by default; the Jarvis proxy and hosted Sysdig Secure do not.
Built by the Epsilon team at Connected Services BC (NR Sector Digital Services).
Docs: setup & usage (this file) · tool catalog & env vars · architecture & security · RFC/RFD playbook
Releases: signed bundled runtimes, verification, and the transitional pinned-source model are documented in Packaging and Releases.
What It Does
RAVEN gives your local LLM direct access to our Atlassian tools via MCP (Model Context Protocol) servers, and includes an autonomous DevOps pipeline that detects production errors, triages them with AI, generates code fixes, and creates pull requests.
| Server | What You Can Do |
|---|---|
| Jira | Search by JQL; read/create/update/comment/transition issues (Epic Link aware); log time; manage watchers, versions, and the sprint lifecycle; link issues; create RFCs, RFDs and RFD-subtasks (custom fields, field metadata) and book deployment-calendar slots |
| Confluence | Search (CQL + filtered); navigate the space tree; read/create/update/delete/move pages as Markdown; manage attachments, labels, and comments |
| Bitbucket | Browse/search code; review PRs end-to-end (diff, comment, approve, merge, decline); commit history and blame; tags; CI build status; create branches, repos, and PRs; commit single files server-side; push local branches (host-pinned, never forced) |
| SharePoint | Search, browse, and read BC Gov SharePoint Online project documentation — sites, document libraries, files (Word/PDF/images), and site pages (read-only) |
| Azure DevOps / TFS | Work items (WIQL), repos and branches, pull requests, and YAML/designer and classic build pipelines on Azure DevOps Services or Server; classic release pipelines on Server |
| Assets (CMDB) | Query the Jira Assets CMDB — apps, environments, tech stacks, people, org portfolios; object/schema introspection |
| Server Monitor | Discover deployed apps; search Tomcat/Apache logs; compare versions; diff configs; JVM heap (read-only, over SSH) |
| IMIS | Search the server inventory and SSH in to browse apps and configs (read-only) |
| Sonar | Code issues, quality gates, security hotspots, project metrics, and local sonar-scanner runs |
| Artifactory | Repositories, artifacts, checksums, properties, AQL searches, build-info, protected transfers, and guarded copy/move/delete operations |
| Jenkins | Inspect and configure jobs; run and monitor builds; manage queue items, artifacts, promotions, and credential metadata/lifecycle |
| GitHub | GHAS security: publish SARIF findings to code scanning; triage code/secret/Dependabot alerts; issues, PRs, and org security posture — all gated by a repository allow-list |
| Sysdig Secure | Sysdig's US-East-hosted MCP server, connected directly by your MCP client with browser OAuth; tools and permissions are managed by Sysdig |
| RFC Buddy | Filter and search RFCs from current/completed schedule; updates/advances the API-side baseline |
| Health | Sprint velocity, issue aging, composite health scores, workload, portfolio comparison |
| Overview | One-shot cross-system project summary (Jira + Confluence + Bitbucket) |
| Bug Classifier | Cluster Jira bugs by shared root cause across projects |
| Jarvis | Query the central Jarvis application inventory (dynamic proxy to a BC Gov API) |
See docs/TOOL_INVENTORY.md for the complete tool catalog, per-server counts, and the read/write split. The Atlassian-backed servers (Jira, Confluence, Bitbucket, Assets, Overview, Health, Bug Classifier) share a single authentication session — log in once, use everywhere. Server Monitor, IMIS, Azure DevOps, Jarvis, Sonar, Jenkins, Artifactory, RFC Buddy, SharePoint, and GitHub authenticate separately via credentials in ~/.raven/.env.
Release paperwork (RFCs and RFDs): there is no separate RFC/RFD server. The Jira tools create the tickets and book deployment slots. Ask for it with /rfd: the Claude Code skill in .claude/skills/rfd/ or the VS Code Copilot prompt in .github/prompts/rfd.prompt.md. Both follow the RFC/RFD playbook. RFC Buddy is a separate tool for searching RFC schedules.
Sysdig Secure is configured in .mcp.json as a remote type: "http" MCP server at https://secure.sysdig.com/mcp/secure. Raven runs no Sysdig code and stores no Sysdig credentials: sign in through your MCP client (in VS Code, choose Sign in for sysdig-secure-mcp in the MCP servers panel) and the client handles OAuth. The available tools and what they may change are defined by Sysdig and limited by your Sysdig role; Sysdig responses do not pass through Raven's PI scrubber. The BC Gov Sysdig account is configured for US East; the completed Security Threat and Risk Assessment (STRA) and data-sovereignty acknowledgement cover that deployment, not other regions. Do not change the endpoint for this account to another region. Copilot cloud agent and code review do not support remote OAuth MCP servers. See Sysdig's hosted MCP documentation.
Autonomous Pipeline (@nrs/pipeline)
A standalone CLI that detects production errors via SSH, triages them with AI, generates code fixes, and creates pull requests — all autonomously. Uses the GitHub Copilot SDK with claude-sonnet-4.6. Supports single run, watch mode, and Jira backlog mode.
See AUTONOMOUS_DEVOPS_PIPELINE.md for architecture and DEVOPS_PIPELINE_USAGE.md for CLI usage.
Server Monitor Web UI (@nrs/server-ui)
A web dashboard for monitoring BC Gov application servers — error counts, log searching, JVM heap, deployed versions, config diffs, deployment history, and alerting. Runs locally at http://localhost:3777.
See SERVER_MONITOR_UI.md for setup and usage.
Search Ranking
Jira search results are scored and ranked by relevance (0-100) using:
- Recency (35%) — recently updated issues rank higher
- Status (25%) — In Progress/Open above Done/Closed
- Content richness (20%) — well-described tickets with labels/components score higher
- Search rank (10%) — Jira's own relevance ordering
- Priority (10%) — Blocker/Critical above Low
Results are grouped by age tier: Current (<1yr), Recent (1-3yr), Legacy (>3yr).
Confluence Ranking
Confluence search results are scored and ranked by relevance (0-100) using:
- Recency (40%) — recently updated pages rank higher (pages go stale faster than issues)
- Search rank (30%) — Confluence's own CQL relevance ordering
- Content signal (20%) — excerpt presence and length as a proxy for page richness
- Title match (10%) — bonus when query terms appear in the page title
Results are grouped by age tier: Current (<1yr), Recent (1-3yr), Outdated (>3yr). The search tool instructs the LLM to always read the top pages via read_pages before summarizing — never just listing search results.
SharePoint
Read-only access to BC Gov SharePoint Online for project documentation, design docs, requirements, and mock-ups stored in document libraries and site pages. All 8 tools are read-only, permission-trimmed to what the logged-in user can see:
search_sharepoint— keyword/KQL search across everything the user can see, ranked with hit-highlighted snippetslist_sites— find SharePoint sites by name or keywordget_site— a site's metadata and its document librarieslist_folder— files and sub-folders of a document library folderget_file_info— a single file's size, modified date, version, and URLread_document— a file's content inline (.docx→ markdown,.pdf→ extracted text, plain text verbatim, images → rendered image)read_page— a modern Site Pages article (.aspx) as markdowndownload_file— save a file to the local protected download directory
See Configure Authentication for the SharePoint login flow and env vars.
FOIPPA note: inlined image content (diagrams, mock-ups, screenshots) returned by read_document may contain personal information visible to the AI and cannot be PI-scrubbed. Text-based SharePoint results use Raven's PI scrubber; the Jarvis proxy and hosted Sysdig do not.
Prerequisites
- Node.js 24 LTS recommended; supported versions are 22.12+ within 22.x, 24.x, or 26+ (check:
node -v). Vitest 5 does not support Node 20, 23, or 25. - Playwright Chromium — install with
npx playwright install chromium; repeat after Playwright dependency updates. - An MCP-capable chat client — see Choosing a Client
- BC Gov IDIR credentials for SiteMinder authentication
Quick Start
1. Clone and Build
RAVEN is a TypeScript monorepo (npm workspaces):
git clone https://github.com/bcgov/raven.git ~/Projects/raven
cd ~/Projects/raven
npm install
npx playwright install chromium
npm run build
2. Configure Authentication
Copy the environment template and fill in your credentials:
mkdir -p ~/.raven && chmod 700 ~/.raven
cp .env.example ~/.raven/.env
chmod 600 ~/.raven/.env
Then edit ~/.raven/.env with your values:
ATLASSIAN_BASE_URL=<base url including https>
ATLASSIAN_EMAIL=<your gov.bc.ca email>
ATLASSIAN_PASSWORD=<your IDIR password>
RAVEN_SCRUB_PI=true
Special characters in values: if a value contains
#(or starts/ends with a quote character), wrap the whole value in double quotes — for exampleATLASSIAN_PASSWORD="p@ss#word". An unquoted#starts a comment and silently truncates the value, which breaks authentication. Quoted and unquoted values are both accepted everywhere.
OS-encrypted storage: instead of the plain-text file, an interactive setup script can store credentials OS-encrypted —
scripts/setup-credentials.ps1on Windows (DPAPI) orscripts/setup-credentials-mac.mjson macOS (login keychain). Both scripts are included underscripts/in platform release bundles. Press Enter to keep a stored value, or type:clearat its prompt (including secret prompts) to erase it. Atlassian fields are optional. Remove an erased value from~/.raven/.envtoo if it exists there; the plain-text file remains a fallback.
Optional Jira setting for Epic Link writes:
# Jira Epic Link custom field ID (default: customfield_10006)
JIRA_EPIC_LINK_FIELD=customfield_10006
Optional Azure DevOps settings:
# Azure DevOps Services (opt in; set the Cloud URL explicitly)
ADO_BASE_URL=https://dev.azure.com
ADO_DEFAULT_COLLECTION=<your Cloud organization name>
ADO_PAT=<your Personal Access Token>
ADO_DEFAULT_PROJECT=<default team project name>
# For Azure DevOps Server instead, set its HTTPS base URL and use a Server collection:
# ADO_BASE_URL=https://your-ado-server.example.com
# ADO_DEFAULT_COLLECTION=DefaultCollection
# Individual tool calls except list_projects can override the organization/collection with `collection`.
# Optional API version override (default 7.1; Server 2020: 6.0, Server 2019: 5.0)
# ADO_API_VERSION=6.0
For Cloud, keep ADO_BASE_URL at the host root; put the organization only in ADO_DEFAULT_COLLECTION (or a tool's collection argument). Supplying it in both places duplicates the path. Work-item comments use the documented preview API on 6.0/7.0/7.1; older API versions do not support comment operations in this client.
Use get_work_item_history with a work item id to read field and relationship changes in revision order. It returns 20 updates by default (up to 100); use skip to page forward. Long individual values are truncated in the tool output. Use add_work_item_comment with id and non-empty text to post a comment; confirm this write action with the user first. Both tools accept optional project and collection overrides.
Classic release definitions on Azure DevOps Services use a separate vsrm.dev.azure.com host; list_release_pipelines supports Server only and reports an explicit error for Cloud rather than switching hosts.
Optional Jarvis API settings:
# Jarvis application inventory; base URL is required when a token is set
JARVIS_TOKEN=<your Jarvis Authorization Token>
JARVIS_BASE_URL=<your Jarvis base URL, e.g. https://jarvis-api.example.gov.bc.ca>
# /mcp is appended automatically if omitted
Optional SonarQube settings (opt in to Cloud with an explicit URL):
SONARQUBE_URL=https://sonarcloud.io
# US Cloud region: https://sonarqube.us (Enterprise); for Server, use its HTTPS base URL instead.
# SONARQUBE_URL=https://sonarqube.example.gov.bc.ca/sonar
SONARQUBE_TOKEN=<your SonarQube user token>
SONAR_SCANNER_BIN=<optional path to sonar-scanner binary if not on PATH; Windows looks for sonar-scanner.bat on PATH>
For SonarQube Cloud, pass the organization key to sonar_list_projects. For scans, the generic scanner accepts sonar.organization in sonar-project.properties or -Dsonar.organization=<key> in sonar_run_scan's extraArgs. For .NET/MSBuild scans (including auto-detected .NET projects), pass /o:<key> in extraArgs instead: SonarScanner for .NET does not read sonar-project.properties, and -Dsonar.organization becomes /d:sonar.organization, not the required /o: option. SonarQube Server does not require an organization.
For US Cloud scans, also set sonar.region=us (for example -Dsonar.region=us in extraArgs); the scanner defaults to EU. Check SonarSource's US-region scanner minimum versions before scanning.
Azure DevOps Services and SonarQube Cloud are external to BC Gov infrastructure. Before enabling them for enterprise data, complete the applicable privacy and data-residency review; SonarQube Cloud currently offers EU and US regions, not a Canadian region.
The Sonar MCP sonar_list_projects tool lists projects the configured user token can browse (key, name, and visibility) within the supplied Cloud organization or on the configured Server, following pagination automatically. It does not require SonarQube administrator permissions or a scanner installation.
Optional Jenkins settings:
JENKINS_URL=<your HTTPS Jenkins base URL>
# Dedicated Basic Auth is preferred for autonomous operation:
# JENKINS_USER=<your Jenkins username>
# JENKINS_TOKEN=<your Jenkins API token>
# Optional protected local directories:
# RAVEN_JENKINS_CONFIG_DIR=~/.raven/jenkins-configs
# RAVEN_JENKINS_DOWNLOAD_DIR=~/.raven/jenkins-downloads
# RAVEN_JENKINS_SECRET_DIR=~/.raven/jenkins-secrets
If dedicated Jenkins credentials are not set, the Jenkins MCP uses cached SMSESSION authentication. It also switches to SMSESSION for the rest of the process if a Basic request is redirected to the SiteMinder login (which means JENKINS_URL points at a SiteMinder-protected host); see packages/jenkins-mcp/README.md. Job configuration updates require an expected SHA-256 and exact XML from the protected config directory. Credential writes accept secret references from environment variables or mode-restricted files; raw secret values are not accepted as tool arguments.
Optional GitHub settings:
# GitHub Personal Access Token. Required scopes: security_events (GHAS tools),
# repo or fine-grained issues/pull_requests/contents (issue and PR tools)
GITHUB_TOKEN=<your GitHub PAT>
# Required: comma-separated repos the tools may target (owner/repo or owner/*)
GITHUB_REPOSITORY_ALLOWLIST=octo-org/octo-repo,octo-org/*
# GITHUB_API_URL=https://api.github.com
# Feature flags — merge and autofix tools are disabled unless explicitly enabled:
# GITHUB_ENABLE_MERGE=true
# GITHUB_ENABLE_AUTOFIX=true
Replace the illustrative octo-org entries with repositories you actually intend to authorize; separate multiple entries with commas. GitHub uses octo-org/octo-repo in its documentation examples.
All GitHub tools refuse to run without GITHUB_REPOSITORY_ALLOWLIST; mutating tools additionally require a confirm: true argument per call.
Optional RFC Buddy settings:
# RFC Buddy Base URL and Personal Access Token
RFCBUDDY_URL=<base URL for RFC Buddy API, e.g. https://rfcbuddy.example.com/api/v1>
RFCBUDDY_PAT=<your Personal Access Token (PAT)>
Optional Artifactory settings:
# HTTPS is mandatory. These dedicated variables intentionally do not fall back
# to ATLASSIAN_* even if both services currently use the same IDIR account.
ARTIFACTORY_URL=<internal Artifactory HTTPS base URL; ask the RAVEN maintainer>
ARTIFACTORY_EMAIL=<your gov.bc.ca email>
ARTIFACTORY_PASSWORD=<your IDIR password>
# RAVEN_ARTIFACTORY_UPLOAD_DIR=~/.raven/artifactory-uploads
# RAVEN_ARTIFACTORY_DOWNLOAD_DIR=~/.raven/artifactory-downloads
# RAVEN_ARTIFACTORY_MAX_TRANSFER_BYTES=536870912
# RAVEN_ARTIFACTORY_DOWNLOAD_TIMEOUT_MS=1800000
# RAVEN_ARTIFACTORY_DOWNLOAD_REDIRECT_HOSTS=<comma-separated approved storage hostnames>
Create transfer directories only when needed and restrict them with chmod 700; upload source files must be regular files with mode 600. Artifactory credentials are attached to requests only inside the local MCP process and sent only to the configured HTTPS endpoint.
Optional SharePoint settings (dedicated cookie-based auth — separate from Atlassian Basic Auth and not shared with Jira/Confluence/Bitbucket):
# SharePoint Online tenant to connect to
SHAREPOINT_URL=https://<tenant>.sharepoint.com
# Optional: default site path used when a tool call omits sitePath
# SHAREPOINT_DEFAULT_SITE=/sites/MyProject
# Optional: how long a cached session is trusted before re-auth (seconds; default 28800 = 8h)
# SHAREPOINT_SESSION_TTL=28800
# Optional: where download_file saves files (default ~/.raven/sharepoint-downloads)
# RAVEN_SHAREPOINT_DOWNLOAD_DIR=~/.raven/sharepoint-downloads
# Optional: reject download_file for files larger than this (bytes; default 536870912 = 512 MB)
# RAVEN_SHAREPOINT_MAX_DOWNLOAD_BYTES=536870912
# Optional: SPO rate-limit tuning (burst capacity / sustained requests-per-second)
# RATE_LIMIT_SPO_BURST=10
# RATE_LIMIT_SPO_RPS=4
Log in once with a browser flow that captures the SharePoint FedAuth/rtFa session cookies:
npx raven-auth --sharepoint
Add --force to ignore the cached pair and log in again. The session is cached and reused across tool calls until SHAREPOINT_SESSION_TTL expires, then it re-prompts. If you already hold valid cookie values (e.g. from another tool), you can skip the browser login entirely by setting SPO_FEDAUTH and SPO_RTFA directly in ~/.raven/.env instead.
When using Jira MCP tools, both create_issue and update_issue accept an optional epicKey parameter to set Epic Link.
Where do I get the ATLASSIAN_BASE_URL? This is an internal BWA (Basic Web Auth) hostname that bypasses SiteMinder. It is not published publicly. Ask the RAVEN maintainer or your team lead for the URL.
Each MCP server loads this file at startup via dotenv — credentials never appear in your MCP client config or get exposed to the LLM. See Authentication for details.
3. Connect to Your LLM Client
Copy this MCP configuration into your client (see Choosing a Client for where to paste it):
{
"mcpServers": {
"jira": {
"command": "node",
"args": ["/path/to/raven/packages/jira-mcp/dist/index.js"]
},
"confluence": {
"command": "node",
"args": ["/path/to/raven/packages/confluence-mcp/dist/index.js"]
},
"bitbucket": {
"command": "node",
"args": ["/path/to/raven/packages/bitbucket-mcp/dist/index.js"]
},
"sharepoint": {
"command": "node",
"args": ["/path/to/raven/packages/sharepoint-mcp/dist/index.js"]
},
"overview": {
"command": "node",
"args": ["/path/to/raven/packages/overview-mcp/dist/index.js"]
},
"health": {
"command": "node",
"args": ["/path/to/raven/packages/health-mcp/dist/index.js"]
},
"server-monitor": {
"command": "node",
"args": ["/path/to/raven/packages/server-mcp/dist/index.js"]
},
"imis": {
"command": "node",
"args": ["/path/to/raven/packages/imis-mcp/dist/index.js"]
},
"assets": {
"command": "node",
"args": ["/path/to/raven/packages/assets-mcp/dist/index.js"]
},
"bug-classifier": {
"command": "node",
"args": ["/path/to/raven/packages/bug-classifier-mcp/dist/index.js"]
},
"ado": {
"command": "node",
"args": ["/path/to/raven/packages/ado-mcp/dist/index.js"]
},
"jarvis": {
"command": "node",
"args": ["/path/to/raven/packages/jarvis-mcp/dist/index.js"]
},
"sonar": {
"command": "node",
"args": ["/path/to/raven/packages/sonar-mcp/dist/index.js"]
},
"jenkins": {
"command": "node",
"args": ["/path/to/raven/packages/jenkins-mcp/dist/index.js"]
},
"rfcbuddy": {
"command": "node",
"args": ["/path/to/raven/packages/rfcbuddy-mcp/dist/index.js"]
},
"artifactory": {
"command": "node",
"args": ["/path/to/raven/packages/artifactory-mcp/dist/index.js"]
}
}
}
Important: Replace /path/to/raven with your actual clone path (e.g., ~/Projects/raven). Credentials and settings (including PI scrubbing) are loaded from ~/.raven/.env — they do not belong in this config file.
4. Configure Server Monitoring (Optional)
The Server Monitor MCP requires two things: a servers.conf file listing the servers it can connect to, and SSH credentials in ~/.raven/.env.
4a. Create ~/bin/servers.conf
This file is the single source of truth for server names and connection details. The Server Monitor MCP will fail to start if this file is missing or empty.
cp servers.conf.example ~/bin/servers.conf
Then edit ~/bin/servers.conf and replace the example hostnames and _A account with your own values. The format is pipe-delimited:
name|hostname|ssh_user|sudo_user|role|description
See servers.conf.example in the repo root for the full format reference including optional apps_base and logs_base fields.
Custom location: If you prefer to keep
servers.confsomewhere other than~/bin/, setSERVER_TOOLS_BIN=/path/to/dirin~/.raven/.env.
4b. Add SSH credentials to ~/.raven/.env
SERVER_A_PASSWORD=<your _A account password>
Note: Quote the password if it contains # or other special characters. Server Monitor requires VPN connectivity to the BC Gov network.
The ssh_user (your _A account) is read directly from servers.conf — you do not need to set it separately in .env.
Raven MCP questions
Does Raven protect personal information?
Yes, Raven-owned servers use PiScrubber to scrub personal information by default, though the Jarvis proxy and hosted Sysdig do not.
Can Raven autonomously fix production errors?
Yes, it includes an autonomous DevOps pipeline that detects production errors, triages them with AI, generates code fixes, and creates pull requests.
How is authentication handled for Atlassian tools?
The Atlassian-backed servers share a single authentication session, allowing you to log in once to use Jira, Confluence, and Bitbucket.