What MisterShell delivers
Verified capabilities at a glance
Remote workers, enterprise sign-in, high availability, and audit export to your SIEM come with the Pro edition; policy engines, session recording, and automation ship with Enterprise; IDS, syslog collection, and external access are licensed add-ons.
✓ Deployment model — Self-hosted — run it as a single self-contained container, or as a full deployment with outbound workers. No SaaS control plane.
✓ Access surfaces — SSH, AWS CLI, Azure CLI, Kubernetes (kubectl), databases (PostgreSQL, MySQL, MariaDB, SQL Server, ClickHouse), interactive RDP, VNC, and web-application sessions — in the browser, or from your own SSH client against the same policy.
✓ Your own SSH client & automation — Point your usual SSH client at MisterShell, sign in with your own SSH key, and work the way you already do — every shell resource you may reach, under the same permissions, policy, approvals and recording as the browser. A full terminal workspace comes with it: the resource tree, several sessions in tabs, and the same in-session AI, without leaving the terminal. Automation connects the same way, so Ansible and the tooling around it keep reaching devices through MisterShell rather than around it. Included in every edition.
✓ External / vendor access — Invite an external vendor or contractor to a single shared session by email — a one-time join link, no account, reached over the Session Proxy add-on. Session policy still applies.
✓ Session policy & per-command control — Firewall-style allow/deny rules — with notify and log flags — at connection time (by location, resource type, tag, role, session type) and per-command ACLs that take a command line apart and require every part of it to be allowed — so a blocked command chained onto a permitted one still never reaches the target. An Approve action on a rule holds the connection until a person says yes.
✓ Approval before access — Session and file-transfer rules can require a human decision before access — the roles you name approve, the access expires on its own, and every request and decision is kept as evidence even after the rule is edited or removed.
✓ Governed file transfer — The location tree doubles as a file catalog, with object stores mounted per location — so where a file lives follows your topology. People move files against the catalog, never the resource directly; an ordered File Transfer Policy decides each copy on location, type, tag, role, direction and both paths — and can require an approval before the transfer starts. The closest worker streams the bytes, so the resource never reaches the store and no new egress path is opened. Every transfer is its own session, with operations and hashes recorded and no file content retained.
✓ Credential storage & vaulting — Encrypted service account credentials with shared or per-user mapping, masked in the UI, plus a personal and team vault for each person’s own logins — shared by role, with history and an audit record of every use — replacing the KeePass file on a share. Rotation and secrets lifecycle stay in your existing secrets system, driven through the REST API or the official Terraform provider; MisterShell does not replace your secrets management.
✓ Identity, SSO & roles — Sign in through LDAP, OIDC, or SAML with groups mapped to roles — OIDC and SAML providers can take group membership from LDAP — and authenticator-app or email MFA for local and LDAP sign-in. Roles start from seeded built-in ones and can be scoped to locations, so a role only reaches the branches of the tree it should.
✓ Session recording & replay — Full recording and replay for terminal sessions and graphical RDP, VNC and web-app sessions — immutable operational evidence in an object store you own (local, S3, or Azure Blob), SHA-256 hashed at capture and verified on replay, with retention set per rule and replay flagging any gaps.
✓ Audit export (syslog/CEF, webhook, Splunk HEC) — Security, policy, API, and AI audit streams exported to your SIEM over syslog/CEF or webhook (JSON or Splunk HEC) — the AI stream carries metadata only, never captured input or output — backed by a policy log of every rule decision and recorded-session timelines.
✓ Health monitoring & config change tracking — Per-metric health history with warn/critical thresholds; configuration snapshots with structured diffs and a timestamped changelog; and fleet-wide Health and Compliance timelines in Review that show when state changed, not every check.
✓ Config push & compliance loop — Author configuration as reusable templates and push it to many resources; Fact Policy checks turn operational facts into pass/fail assertions; and playbooks close a drift → detect → remediate loop, pausing for a human Approve step before the push if you want one. A fleet Compliance timeline records each verdict with the definition it was checked against.
✓ IDS & syslog ingestion — Optional licensed add-ons: passive IDS sensors stream intrusion-detection alerts, and collectors ingest device and server syslog — both attributed to the right resource and location.
✓ Embedded automation (reacting to events, not replacing your runbooks) — Built to react to what happens in your estate and to hand it on — run an AI agent over a config diff or a failed check, raise a ticket, call a webhook, email an owner, generate a report — with a visual canvas for event-driven and scheduled playbooks, Switch steps that branch on conditions, and an Approve step that pauses the run for a human decision. It is not here to replace Ansible or your existing runbooks: keep those, and point them at MisterShell so they reach devices on the same governed path as your engineers.
✓ AI assistance (BYO LLM, governed) — In-session AI assistant grounded in operational context; bring your own LLM from ten providers (Anthropic, OpenAI, Ollama, Azure OpenAI, Google, Mistral, xAI, Cohere, OpenRouter, AWS Bedrock). AI guardrails keep it read-only by default, and every chat, agent run, model request, and tool call lands in one AI audit trail. Agent platforms you already run can act through MisterShell as a user, with tokens from your own OIDC identity provider — no personal API keys to hand out.
✓ Network model — Outbound-only workers deployed inside protected zones — no inbound firewall rules required — with strict SSH host-key verification applied to live sessions and background checks alike.
✓ Inventory, RBAC & session policy as code (Terraform / MCP) — The official Terraform provider goes past inventory: locations, resources, tags and credentials, but also roles, users and role grants, session policy rules and per-command ACLs, identity providers and their mappings, workers, log destinations, settings, and AI models, prompts, agents and skills — so who may reach what, and what they may run, lands in a reviewed pull request rather than a console. Recording, fact and config policies and automation playbooks are authored in the UI, not yet in the provider. Plus a REST API for scripts and CI, and a built-in MCP endpoint giving external AI access to inventory, changelog, and diagnostics with exactly the caller's permissions, read-only by default.
