The vocabulary, defined plainly.

Software operating partnership

A firm that builds, hardens, and operates the custom software a business runs on, under a recurring engagement. Not a dev shop (which builds and leaves), not an MSP (which watches infrastructure but not the application), not staffing (which delivers labor without accountability). Keepstone is one.

Software Ops

The recurring work of running a piece of software well: monitoring, incident response, maintenance, releases, documentation, and stakeholder reporting. Distinct from building software. Sold by Keepstone in six fixed monthly tiers.

Software Ops Agent Framework

Keepstone’s proprietary set of agent-native layers wrapped around every system we operate. Eight layers: Engineering Workflow, Infrastructure & DevOps, Observability, Triage & Support, Documentation & Training, Security, Business Continuity & DR, and Governance. Subagents and skills do most of the work; operators make the named decisions.

Agent-native

Designed from day one to run with AI subagents and skills as primary actors, not as bolt-ons to a human-driven process. Different from "agent-assisted," where humans do the work and an LLM helps; agent-native flips it — agents do the work and operators make the calls.

Vibe-coded system

Software built with AI coding tools (Lovable, Bolt, Replit, v0, Cursor, Claude Code, and similar) by a non-engineer or by a builder working without traditional engineering rigor. Often works. Often runs the business. Almost always fragile in production until it’s hardened.

Key-person system

A piece of custom software the business depends on, but only one person knows how it works. Common shape: a partner, staff member, or long-time contractor built it. No documentation, no second set of eyes. Institutional risk that wants to be turned into an institutional asset.

Workflow-as-system

A process the business already runs reliably (in spreadsheets, email, a SaaS tool, or someone’s head) that should be a real piece of software. Discovery turns it into one.

Framework-compatible

A state in which a system meets the operational baseline required to be run inside the Keepstone Software Ops Agent Framework. Source control with reviewable changes, separated environments, secrets out of code, proven backups, observability wired, deploy pipeline, minimum auth hygiene, current documentation. Hardening is the work of getting there.

Harden

The one-time engagement that brings an existing tool up to operable standard before it goes into Keepstone Operations. Fixed price, scoped from the Assessment. Tiered $5k–$55k.

Assessment

Free two-week structured review of an existing system, scored against the Framework. Eight-category technical review, plain-English risk summary, written report, fixed quote for Harden plus Operations. No obligation to engage.

Discovery

$3,000 fixed-scope engagement for net-new builds. Working sessions, technical discovery, a written specification, and a fixed quote for Build plus Operations. Fee credited in full against Build if the engagement proceeds.

Build

Net-new custom software built from a Discovery spec. Modern agent-native stack. Operational standards shipped from day one — source control, environments, observability, framework compatibility — with a handoff into Operations included.

Operate / Operations

Keepstone’s recurring monthly service. The six standing disciplines applied to your system every month: Monitor, Respond, Maintain, Release, Document, Report. Sold as one fixed monthly number per tier.

Tier

A six-level scale (Tier 1 through Tier 6) used to set the price and operational envelope for an engagement, derived from business criticality, workflow complexity, integration count, data sensitivity, and fragility. The same tier is used across Harden, Build, and Operations.

High-availability rider

Optional add-on available on Tiers 4–6 of Operations. Adds 24×7 pageable response, tightened RPO/RTO targets, and quarterly failover drills.

Trivial enhancement

A change so small that an agent can simply do it under guardrails — copy tweaks, field additions, small UI changes, minor enhancements. Included in Operations under your monthly fee, no ticket-and-quote dance.

Enhancement sprint

A scoped block of capacity for changes too big to absorb under Operations. Written scope, fixed price, agreed ship date. Several pricing formats depending on cadence.

Subagent

A scoped AI agent within the Framework dedicated to a specific class of work — for example, a triage subagent that classifies incoming issues, or a docs-sync subagent that updates documentation against the code. Built on top of foundation models (Claude, GPT-4, etc.) with role-specific guardrails.

Skill

A reusable capability a subagent can invoke — for example, a "Lovable deploy" skill, a "Supabase backup verify" skill, an "incident post-mortem draft" skill. Skills are how the Framework acquires platform-specific competence.

MCP tool

An external system exposed to a subagent via Anthropic’s Model Context Protocol — for example, a GitHub repo, a Linear project, a database, a monitoring dashboard. MCP tools are the connective tissue between subagents and the systems we run.

RPO / RTO

Recovery Point Objective and Recovery Time Objective. RPO: the maximum acceptable amount of data loss measured in time (e.g., "we can afford to lose up to one hour of data"). RTO: the maximum acceptable downtime after an incident (e.g., "we must be back up within four hours"). Set per system, in writing, in the engagement letter.

Standardized observability

A consistent set of monitoring, logging, error tracking, and alerting practices applied across every Keepstone-operated account, regardless of the underlying stack (Lovable, Replit, Bolt, Supabase, Vercel, AWS, etc.). Means an operator can pick up any account without learning a bespoke setup.

Governance layer

The Framework layer that decides fit, enforces boundaries, defines required hardening, and keeps systems inside operable conditions. The one layer that stays firmly in operator hands, because fit and scope are judgment calls that need a named human behind them.

Operator

In Keepstone context, a named human at our firm responsible for an engagement — fit decisions, architecture, incidents with business weight, the handful of changes that need a person behind them. Distinct from agents (which do the bounded work) and from "engineers" in the dev-shop sense.

Engagement letter

The written document every Keepstone engagement runs on. Names the principal, the scope, the price, the tier, any riders, the support/enhancement boundary, and the 30-day exit clause. Updated when scope changes; never an open-ended retainer.

Missing a term? Tell us at hello@keepstone.tech and we’ll add it.