Watch the four-minute Cowork demonstration.
The visual above explains the underlying system. Anthropic's walkthrough shows what this style of work looks like inside a familiar AI workspace.
Watch the demo// How it works
We run on this too.
Sidekick builds an AI operating system around the way you already work. For one person or a whole organization, it connects the right context and tools, carries repeat work through a reliable process, and stops when your judgment matters.
// The systemSee the interface See the possibilities Try it yourself
Your tools feed one operating system. Useful work comes out.
Think of the AI interface as the front door. The operating system behind it remembers the job, moves it to the right place, checks what comes back, and knows when to knock on your door for a decision.
The tools you already use
EmailCRMCalendarTeams + chatFilesForms + intake
ReadySidekick operating systemIt keeps the useful context, sends each job down the right path, and knows where your approval begins.
01Email + follow-up
- Inbox triagesorts what needs attention
- Reply draftsprepares responses in your voice
- Commitmentskeeps promises from getting lost
02Meetings + decisions
- Meeting prepbrings the useful context
- Notes + recapcaptures what matters
- Next actionsturns decisions into follow-through
03Sales + relationships
- Pipeline watchflags deals going quiet
- Prospect prepbriefs you before calls
- Nurture promptskeeps relationships moving
04Reporting + operations
- Client healthspots risks and openings
- Weekly reportsassembles the real picture
- Knowledgekeeps useful work findable
You stay in controlSending, publishing, spending, and consequential changes wait for your approval.
✓Checked work, ready for review
Verified corrections make the next comparable run better.
What can AI take off your plate?
Explore practical examples before opening the technical details.
See what AI can doStart with the free Starter Kit.
Use a smaller version of the same operating idea to complete a useful first piece of work.
Get the Starter Kit// Complexity, compressed
One simple trigger. A disciplined system behind it.
One-Shot is a current Codex workflow we use inside Sidekick. You describe the job in ordinary language. The system carries it from a clear brief to a checked result without asking you to remember or manually coordinate every stage.
Current Codex implementationOne trigger, eight visible stages
You say
“One-shot this.”
Add the work itself, the outcome you want, and any context you already have.
The system hearsOwn the whole job.
Frame it, find what matters, build it, challenge it, verify it, and return it with proof.
What runs behind the phrase
The route adapts to the job. Research and specialist tools run only when they add value.
01Understand
Frame the job
Turn the request into a clear outcome and stopping point.
OutcomeAudienceBoundariesSuccess bar
02Understand
Load the context
Bring forward only the state and sources that matter now.
Current statePrior decisionsRelevant files
03Decide
Research if needed
Resolve current, uncertain, or contested facts before they shape the work.
Primary sourcesCounter-evidenceOpen questions
04Decide
Challenge the plan
Find weak assumptions, failure points, and approval gates before building.
RisksAssumptionsTestsAuthority
05Build
Use the right craft
Route the job through the specialist workflow that fits the deliverable.
WritingWebAnalysisConnected tools
06Build
Verify directly
Check the actual result with evidence that matches the work.
Source checksTestsRendered proofAccessibility
07Prove
Improve what matters
Fix material defects, then stop when another pass would only add noise.
Verified defectsHigh-confidence fixesTwo loops max
08Prove
Hand off with proof
Return the result with evidence, assumptions, risks, and the exact next action.
Finished workEvidenceOpen gatesNext action
Loop 01↺
Research loop
Uncertainty returns to evidenceNothing decision-changing moves forward as an unsupported fact.
Loop 02↺
Improvement loop
Defects return to their sourceThe system fixes the stage that produced the problem instead of polishing around it.
Loop 03↺
Verification loop
Every major handoff must passResearch, planning, building, and final delivery each carry their own proof.
Human judgment stays at the boundarySending, publishing, spending, and consequential changes still wait for you.
The operator has to remember the machinery.
- 01Clarify the brief
- 02Find the right context
- 03Research current facts
- 04Build and challenge a plan
- 05Choose specialist methods
- 06Review intermediate work
- 07Correct material defects
- 08Run final QA
The machinery is encoded into the trigger.
“One-shot this.”
You provide intent, context, and judgment. The system provides the operating discipline, remembers the quality gates, and shows its work at the end.
Less operator loadVisible processChecked result
What this does and does not promise
A simple trigger gives a less experienced operator access to the same disciplined process. The final result still depends on the request, available context, tool access, evidence, and human judgment.
The same idea in everyday work
A meeting ends. The follow-through is already moving.
- 01Notes and context arrive
- 02Follow-through is prepared
- 03Names, dates, and claims are checked
- 04You approve what leaves the business
// One system, two scalesFor one operator For an organization
The same operating foundation for one person or a whole organization.
The responsibilities stay stable: understand the work, keep the right context, use the right tools, verify the result, and preserve human control. The depth of governance grows with the number of people, systems, and risks involved.
One system, two scales
Simple enough to use. Structured enough to trust.
Solo
A personal AI workspace built around your role, priorities, voice, and repeat work. You learn to run it independently, with support while the habits and workflows settle in.
Explore SoloFractional
The same foundation across a business, with governance, access controls, risk management, audit logs, adoption support, and ongoing ownership across people and workflows.
Explore Fractional// Honest boundaries
Reliable work still needs clear routes, receipts, and review.
The system is built around the tools and permissions that are available today. Important recurring work keeps a receipt and a recovery path, and consequential actions remain under human control.
Natural-language routing is not perfect
Named workflows and clear trigger phrases remain more reliable than hoping the platform infers the right process from every casual request. The system makes the reliable path easy and keeps a manual route visible.
Scheduled work still needs supervision
A recurring task can time out, lose tool access, or return no useful finding. Each important run needs a receipt, a missing-run path, and a clear owner for repair.
Connector capabilities are uneven
Read, write, edit, and sharing behavior differs by tool and platform. Every workflow is designed around the operations that are actually available, not the ones a connector name appears to promise.
The platform moves underneath us
Microsoft, Claude, ChatGPT and their connectors evolve quickly. Versioned workflows absorb most change, but occasional retesting and repair are part of operating the system honestly.
// The full system
Open the parts you want to inspect.
The visual above is the working shape. These chapters preserve the complete architecture, quality controls, evidence, limits, and delivery process without forcing everyone to read the technical version first.
The architecture is evergreen. Dated operating snapshots, measured results, and platform limits are labelled. Every build uses five operating layers, with an evidence loop that carries verified lessons back to the smallest surface that should change.
01Ground rules
Identity, authority, sources, safety, and what must always hold.
02Context + state
Current truth, durable task resumes, decisions, relationships, and work history.
03Orchestration + skills
Briefs, plans, specialist routing, safe work graphs, and repeatable workflows.
04Live + scheduled work
Connected tools, monitors, briefs, dashboards, and recurring operating cadences.
05Verification + approval
Direct proof, quality gates, staged outputs, and human control of consequential action.
Evidence loopObserve → test → improve → check again
Run receipts, feedback, governance decisions, and evaluations help decide what deserves to become a rule, workflow, connection, or schedule.
Five operating layers, one governed loop.
01What it solvesAI workspaces decay by default
Adopt Claude or ChatGPT seriously and the first few weeks feel like progress. Then the slide starts. The same failures show up across very different people, in the same order, and they compound quietly enough to look like the model getting worse. It does not get dumber. It gets buried.
01
Memory is shallow and uncurated Without the system
The big tools carry facts across sessions now, which is exactly why the first weeks feel great: it starts to know you. But the memory keeps whatever passes through, with no sense of what is current, what matters, or what should have been dropped. It pulls some things forward and quietly misses others, and the gaps surface right when you had started to trust it.
02
Instructions pile up and contradict Without the system
Most setups grow by appending. New priorities sit next to old ones, current facts beside stale ones, and nothing is ever removed. The file meant to make the assistant smarter becomes the thing making it unreliable. The longer you use it, the worse this gets.
03
Output sounds like the model, not you Without the system
A paragraph describing how you write produces something generic with shorter sentences. A description of your voice is not your voice, the same way a new hire handed your LinkedIn page could not sign your emails. You edit everything heavily or stop delegating writing at all.
04
Usage plateaus at chat Without the system
With no named workflows, trigger phrases, or first win, usage stays at ad-hoc questions. The recurring work that would justify the tool never moves onto it.
The compounding cost
None of these is dramatic on day one. By week six the workspace is a junk drawer, you trust it less than at the start, and the honest read is that the tool never paid back its own learning curve. That is what turns “it knows me” into “it has gotten worse.”
Want the plain-English version of this? Read Why Your AI Gets Worse the More You Use It.
02What you receiveYou get a workspace that gets better the more you use it
The deliverable is an operating environment that gets more useful with use: context compounds, recurring work becomes a named workflow, and verified improvements propagate forward into new builds. The same operating pattern runs in our ChatGPT workspace and in Claude + Cowork client builds. The exact files, connectors, and interface follow the platform and the work.
Each build starts from the current Sidekick foundation and is populated with your context, boundaries, and highest-value workflows. The core is file-first, inspectable, and portable. Your business context stays in a workspace you can read, own, and move; live connectors are added where they make the work more useful.
Your workspace/
├── CLAUDE.md the instruction layer; loads at every session start
├── Start_Here.md one-page entry point, names your first-win task
├── Your_Workflows.md your 2-3 highest-priority workflows, copy-paste prompts
├── INDEX.md master map of every file in the workspace
├── Starter_Prompts.txt the full skill routing table: trigger phrases
├── Skills/ 8 day-one skills + Skills/Operator/ (30-skill library)
├── Guides/ branded PDF guide set, dashboard template, Your First Month
├── Reference/ Business + Comms profiles, Current_State, Action_Items, Decisions
├── People/ one context file per key relationship
├── Meetings/ structured call notes, one file per meeting
├── Work/ durable task resumes, evidence, reviews, and exact next actions
└── SOPs/ · Active/ · Deliverables/ your processes, work in progress, finished output
├── CLAUDE.md the instruction layer; loads at every session start
├── Start_Here.md one-page entry point, names your first-win task
├── Your_Workflows.md your 2-3 highest-priority workflows, copy-paste prompts
├── INDEX.md master map of every file in the workspace
├── Starter_Prompts.txt the full skill routing table: trigger phrases
├── Skills/ 8 day-one skills + Skills/Operator/ (30-skill library)
├── Guides/ branded PDF guide set, dashboard template, Your First Month
├── Reference/ Business + Comms profiles, Current_State, Action_Items, Decisions
├── People/ one context file per key relationship
├── Meetings/ structured call notes, one file per meeting
├── Work/ durable task resumes, evidence, reviews, and exact next actions
└── SOPs/ · Active/ · Deliverables/ your processes, work in progress, finished output
A few root files carry the day-one load: Start_Here.md (how to start a session and your named first win, no placeholders), Your_Workflows.md (your priority workflows with copy-paste prompts), and Starter_Prompts.txt (the routing table that makes triggering reliable). A reader who opens the folder cold finds an entry point, a map, and no internal build artifacts.
03Memory that compoundsThe index is what makes memory reliable
Most people run their AI memory like a junk drawer: everything goes in, nothing comes out, and eventually you are afraid to open it. The fix is a handful of boring disciplines, none of which need a better model. In a Claude Code implementation, CLAUDE.md is the project instruction layer loaded into sessions, so it can carry the identity, routing, memory, and safety rules that must always hold. Other environments use an equivalent instruction layer. [R4]
Give the workspace an index. A simple map of every file and what it holds lets the model look up the one relevant, current thing and read just that. With no index, it rereads the entire filing cabinet to answer one question, fills its working memory (the context window) with noise, and forgets the answer. People read that as the model getting forgetful. What is really happening is it has more memory than it can sort, so it holds all of it at once and weights none of it. The index is also where relevance and recency get decided. For most people hitting the “it cannot remember anything” wall, this is the highest-impact fix almost nobody has in place.
Rewrite the state, do not append to it. The current-state file is rewritten at the end of each session so it always reflects what is true now, instead of stacking session recaps until it contradicts itself. This single discipline kills most of the instruction-pileup decay on its own.
Route memory by type. What is true now (priorities, business, people) lives in one place. What happened when (decisions, meeting notes) lives in another. How things get done (standing rules, workflows) lives in a third. Every “remember this” is sorted to the right type before it is written, so the assistant can actually find it later.
Keep long work resumable. A substantial job gets a compact control surface with its objective, completed milestones, evidence, failed paths, approval boundaries, and exact next action. If a conversation is interrupted, the system resumes from the record instead of reconstructing the job from chat history.
Why this matters
Every drafting and analysis skill reads this layer before producing anything, which is why output quality scales with workspace age here and degrades with it elsewhere. The architecture is the moat; the prompts are replaceable. A weekly scheduled lint pass catches stale content, contradictions, and dead references before they mislead a session.
04Skills + orchestrationSkills do the work. Orchestration makes them one system.
A skill is a saved specification that wraps a recurring piece of judgment: what to read, what to produce, what never to do. A core set is active from day one, and a deeper library opens up as your usage matures, so most builds end up running more than 40 skills in total. Every skill carries explicit trigger phrases, because manual triggering is the reliable path on this platform (more on that under limits).
For substantial work, orchestration sits above the individual skills. It locks the outcome, chooses the specialists, preserves state, and checks each handoff before the next phase begins. Work with real branches can use a small work graph so independent parts run safely and rejoin at one verified result. Simple work stays simple.
The rule that prevents polished failure
Research must be sound enough to plan from. The plan must be sound enough to build from. Novel work proves the smallest complete path before broad production. When a premise breaks, the job returns to the stage that owns it instead of spending a final polish loop on the wrong thing.
| Day-one skill | What it does |
|---|---|
| session | Opens and closes the session: reads context, then updates current state, logs decisions, and integrates corrections. The ritual that makes everything else persistent. |
| daily-brief | A sub-3-minute morning orientation from priorities, action items, decisions, meetings, and live calendar, email, and CRM where connected. |
| post-call | Raw notes or a transcript in; structured meeting notes, signals, a follow-up draft, and a next action out, in one pass. |
| quick-capture | “Remember this” routed to the right file and confirmed in one line. No questions, no analysis. |
| action-items | The live task ledger: add, review, filter by person, prioritize, hygiene sweep. The owner column drives the dashboard split. |
| meeting-prep | A one-page brief for an upcoming meeting: agenda, objectives, relationship context, recent signals. |
| executive-comms | Drafts a message to a named recipient in your calibrated voice and their register. Never sends without approval. |
| dashboard-render | Renders the live daily dashboard from your action items, decisions, and current state. |
| Operator library · introduced as usage matures | |
| Documents | content, docx / pptx / xlsx, outreach, prompt-sharpener: long-form writing, branded documents, and drafts in your voice. |
| Planning | plan / build, decision-brief, strategic-analysis, board-prep: multi-step work spec-locked, scored, and steelman-first. |
| Orchestration | one-shot, brief locking, work graphs, durable resumes, verification, and approval-ready handoffs for substantial jobs. |
| Voice | my-voice / voice-extractor, stakeholder-comms, people-intelligence: voice pulled from real samples, multi-audience handling. |
| inbox-organizer, email-drafter: triage plus reply drafting staged in Gmail. Never sends. | |
| Meetings | meeting-intelligence, week-ahead / weekly-review: transcript processing and the weekly cadence. |
| System health | feedback-loop, sop-capture, tune-this, quality-loop, workspace-eval, context-lint, update-diff, monthly-review, schedule. |
Custom skills, scoped to you
The baseline: we turn your two or three highest-impact workflows into custom skills, with trigger phrases rich enough to catch the natural ways you would ask. From there it grows with you. Once those first workflows are running, the coaching tends to surface the next ones, and clients who take the coaching often finish with half a dozen or more custom skills, covering ground they did not know could be automated when they started, until much more of their operation runs on tooling built specifically for them. Client identities, workflow records, and implementation details remain private.
05Your daily surfaceScheduled work keeps the system useful between conversations
The dashboard is your persistent operating view, rendered as a live artifact from your working files, not a separate database that drifts. Tiles come from your action items, decisions, and current state, and the owner column splits work into “the system can handle” and “only you.” It is plain, inspectable HTML that runs inside your workspace, with a setup guide for adding tiles.
Scheduled tasks handle the work that should not depend on you remembering to ask: morning orientation, follow-up detection, pipeline and delivery scans, context health, and bounded system reviews. Each important run leaves a compact receipt, including when nothing useful was found or when access failed, so silence does not look like success.
Daily
Orientation + follow-up scan
Priorities, action items, calendar, inbox, and relationship signals become a short brief plus only the follow-ups that need attention.
Weekly
Operating + context health
Scans for work that can be moved forward, checks that current state and profiles agree, and flags missing runs or stale assumptions.
Monthly
Evidence review
Mines learning signals and scheduled findings for repeated causes, then proposes the smallest testable improvement or recommends no change.
Honest boundary
Scheduled work depends on the permissions, connectors, and runtime available to it. Long runs can fail, and a schedule is not the same thing as an event-triggered workflow. The system records missing or failed runs and keeps consequential action human-gated. It is an operating cadence, not an unbounded autonomous fleet.
Sidekick
Illustrative Personal WorkspaceLive
⚠ Needs attention
12 days
Project Pine proposal still not sent
The draft is ready for review. A person still needs to approve and send it.
this week
Project Cedar review is due
The next review is in three weeks. This conversation needs a human owner.
Active Projects
8
illustrative
Monthly Target
80%
fictional progress
Proposals Out
2
sample count
Closed This Week
1
sample result
✓ System can handle · 3
1
Draft Project Pine follow-up · 12 days overdue
overdue
2
Update client summary for review call
today
3
Draft renewal notices · 2 clients
this week
★ Only you · 3
1
Send Project Pine proposal · draft ready
12 days
2
Project Cedar review call · due in 3 wks
this week
3
Scope call · Project Birch waiting
today
// Sample outputFictional example showing the structure only. It contains no client or internal operating data. A personalized dashboard uses your approved data, tools, and branding.
06Built safe, proven by measurementA quality floor that has itself been tested
You will paste emails, transcripts, and documents from strangers into the workspace. The safety layer assumes that. External content is treated as data, never instructions: it cannot invoke a skill, change a rule, or authorize an action, and instructions found inside content are surfaced, never silently obeyed. Every outbound skill drafts only; sending is always a human click, which caps the blast radius of any manipulation at a staged draft.
No workspace ships on a feeling. Every build passes eight mechanical floor gates, each a literal pass or fail with no evaluator discretion: cross-client contamination, stale pricing, placeholders shipped as real, wrong identity, leaked internal content, dead references, core-skill integrity, and structural completeness. A single failure is a red verdict that blocks delivery. Then ten graded dimensions score the build against a written bar, ending with an experience score taken after a four-scenario simulation of your first week.
Substantial jobs use the same discipline while they run: challenge the handoff between research, plan, and build; test the first meaningful end-to-end path; verify the exact destination; and keep inaccessible human, legal, safety, or live-system evidence as an open gate instead of averaging it away.
Green · ship
8/8 + all 10 at 7+ + experience 8+
Every gate passed, every dimension strong, zero critical fixes.
Yellow · fix then ship
Gates pass, scores dip
A dimension at 5-6 or experience at 6-7. Fixed and re-run before delivery.
Red · rebuild
Any gate fail
Any floor-gate failure, any dimension at 4 or below, or any critical fix present.
A quality check you have never tested is just a vibe. So we tested ours. In June 2026 we ran a seeded-defect audit against our own framework: ten known defect classes planted into workspace copies, then blind evaluators who did not know what was planted ran the gate.
01
First measurement: 7 of 10 detected honest result
The gate missed stale pricing inside an HTML file, a gutted core skill, and stripped frontmatter. Those classes had relied on evaluator vigilance where they needed mechanical checks.
02
Rebuild: mechanize what vigilance missed
The eight floor gates were built directly from the misses, moving every miss-prone class from graded judgment to grep, diff, and file tests.
03
Re-measurement: 10 of 10, reproducible verified
Fresh blind evaluators, fresh seeds: 10 of 10 detected, zero false reds on compliant builds, scores reproducible across independent runs.
07How it improves over timeThe logs are evidence. The improvement is earned.
The evidence loop captures material corrections, verified misses, recurring friction, useful scheduled findings, and demonstrated successes. Ordinary completion and unsupported self-critique do not qualify as learning signals.
Signals are clustered by their first shared cause. A direct correction to a durable boundary or one high-risk failure can justify an immediate fix; ordinary signals need repetition. Every promoted change starts with a hypothesis, success bar, direct test, rollback path, and a later outcome check. Weak evidence can correctly produce no change.
Illustrative evidence ledger
Fictional records · no client or internal dataDifferent logs answer different governance questions.
Run receipt
RUN-EXAMPLE-042Four sources checked. One unsupported claim removed before review.
Result staged for human approvalOutcome + evidence + open gateFeedback log
FB-EXAMPLE-017The summary missed a delivery constraint named in the brief.
Cluster with comparable missesCorrection + source + first causeGovernance log
GOV-EXAMPLE-009Requested read access to one named calendar event.
Approved once, scoped to that eventRequest + scope + effect + reversibilityChange + evaluation
CHG-EXAMPLE-006Add a required constraint check before drafting begins.
Promote only after the seeded test passesHypothesis + test + approval + rollbackAccess requests are recorded with the requested scope, decision, expected effect, duration, and reversibility. A repeated governance issue can become an improvement signal, but the log never grants authority by itself.
Evidence guide
Which AI records can you trust?
This ledger shows Sidekick's operating pattern. The full guide explains which records can prove activity, authority, and outcomes, plus the privacy, retention, and compliance limits of logging. Read AI Audit Logs
01Observe
Capture a real correction, verified miss, recurring friction, or demonstrated success.
02Record
Write one compact, source-linked signal instead of preserving a transcript.
03Cluster
Group evidence by the first shared cause, not by every downstream symptom.
04Promote
Choose the smallest durable change the evidence actually supports.
05Verify
Test the changed behavior against a success bar and keep a rollback path.
06Prove later
Check the next comparable uses to learn whether the change held, regressed, or should be removed.
What propagates
A verified lesson is wired into the smallest owning surface: current context, a stable rule, a skill, a connector, or a scheduled task. New builds start from the current foundation. Delivered client workspaces are never silently retro-edited; updates remain visible and adoptable.
08Portable by designOne operating kernel, adapted to each environment
The durable architecture is not a Claude folder or a ChatGPT feature. It is the contract between identity and authority, context and state, workflow specifications, tools, orchestration, evaluations, approvals, observability, and controlled improvement. Each platform gets an adapter for how those responsibilities are actually implemented.
Portable operating kernel
Identity + authorityContext + stateWorkflow contractsTools + orchestrationEvaluations + approvalLogs + improvement
Claude + Cowork
Project instructions, workspace files, skills, MCP or connectors, and scheduled work.
Client-delivery patternChatGPT + Codex
Project rules, skills, apps, task state, approvals, logs, and recurring work.
Sidekick operating patternMicrosoft Copilot Studio
Instructions, knowledge, tools, flows, human review, testing, and tenant governance.
Adapter mapping; validate per tenantCustom API system
State store, tool schemas, orchestration, traces, evaluations, approval gates, and deployment controls.
Reference architectureHermes or open source
Skills, memory, toolsets, schedules, model routing, and runtime-specific observability.
Adapter mapping; validate per runtimePortable does not mean identical
Instruction loading, memory, tool permissions, scheduling, tracing, deployment, and governance differ by platform. The adapter is tested against the real environment before a workflow depends on it. OpenAI documents agent tracing and handoffs; Microsoft documents tools, flows, human review, testing, and governance; Claude and Hermes expose project memory or reusable skills in their own ways. [R1-R6]
09Real gaps and limitsCurrent constraints and operating limits
These are the operating constraints we design around as of July 2026. The architecture choices only make sense against them, which is why they are in this document rather than hidden under it. Exact connector and platform behavior is rechecked before a workflow depends on it.
Natural-language routing is not perfect
Named workflows and clear trigger phrases remain more reliable than hoping the platform infers the right process from every casual request. The system makes the reliable path easy and keeps a manual route visible.
Scheduled work still needs supervision
A recurring task can time out, lose tool access, or return no useful finding. Each important run needs a receipt, a missing-run path, and a clear owner for repair.
Connector capabilities are uneven
Read, write, edit, and sharing behavior differs by tool and platform. Every workflow is designed around the operations that are actually available, not the ones a connector name appears to promise.
The platform moves underneath us
Microsoft, Claude, ChatGPT and their connectors evolve quickly. Versioned workflows absorb most change, but occasional retesting and repair are part of operating the system honestly.
What the quality gate does not certify:
- Your activation. The gate certifies the workspace as delivered; whether you embed it in daily operations is coached, not guaranteed.
- Long-term drift. Priorities shift. The lint and review cadences catch drift, but only if you run them or keep the schedules on.
- Voice under pressure. A voice calibrated from polished writing may differ in a tense thread; the review mode converges it over time.
- Connector uptime. The gate cannot test the real-world stability of your Gmail, Drive, or CRM connections.
Fit boundary
The system is wrong for buyers who want fully hands-off AI, who do not work at a computer daily, or who need deep enterprise integration. If that is you, we will say so in the first conversation rather than sell past the line.
10How delivery worksSigned engagement to independent operator
01
Discovery
A live call captures your role, daily tasks, tool stack, and highest-value workflows. No intake forms; the call is the input, and its notes are what the quality gate later traces every claim back to.
02
Spin-up
A skill copies the Master Template and populates your instruction layer, profiles, current state, and priority workflows. Pending improvements are applied first, so every build starts at the current edge.
03
Customization
Your two or three highest-impact workflows are built into custom skills (more if you want to go deeper), voice is calibrated from real samples where provided, and your first-win task is named so day one has a concrete outcome.
04
Quality gate
Floor gates, dimensions, and the first-week simulation. Nothing ships below green.
05
Delivery
You receive the workspace with guides, a named first win, and setup instructions. Async by default, with a live session where fit warrants.
06
Onboarding to independence
Five staged stages pace you from first task to running novel work unaided. Coaching, where purchased, ends on a competence check, not a session count.
What is special, in one paragraph
Plenty of people will configure an AI workspace for you. The difference here is the system around it: context engineered against decay, orchestration that returns errors to the stage that owns them, scheduled work with durable receipts, a quality gate that has itself been measured, an evidence loop that tests improvements before keeping them, and a written record of the limits. The result is an operating environment with a maintenance discipline, documented here in full.
11References + provenanceClaims should be inspectable
External references below document the platform primitives used in the portability map. Sidekick operating claims are labelled with a date and come from our own measured system. Illustrations on this page use fictional records; raw client records, private operating logs, access history, credentials, and private repository paths are not published.
- R1OpenAI Agents SDK: Tracing
Traces, spans, tool calls, handoffs, guardrails, custom events, and sensitive-data controls.
- R2Microsoft Copilot Studio overview
Instructions, context, knowledge, tools, triggers, orchestration, flows, and human review.
- R3Microsoft Copilot Agent Kit
Testing, rubrics, governance, compliance, and repeatable agent quality practices.
- R4Claude Code: project memory
Project instruction and memory loading across workspace scopes.
- R5Model Context Protocol
An open standard for connecting AI applications to external systems, tools, and context.
- R6Hermes Agent: skills system
Reusable skills, external skill directories, and portable skill structure in an open-source agent runtime.
Evidence boundary
Official documentation proves that a platform exposes a capability. It does not prove that a Sidekick implementation uses it successfully. That requires a dated build, direct tests, receipts, and the real user journey to pass in the named environment.
// Your turn
Build a system around the work you repeat.
Same engine, your business. Different skills, your tools, your voice, shaped to how you actually work. Tell us what your week looks like and we will show you what it could run on.