Notion Custom Agents vs win.sh: Two Operating Models
Compare Notion Custom Agents and win.sh on document context, company signals, human authority, local execution and usage budgets.
Founder, Task Machine
Notion Custom Agents and win.sh pose an operating-model choice: do you want shared agents working from pages and databases, or one company operator turning business metrics into projects? A founder evaluating a weekly business review needs to settle who turns a changed number into the next piece of work. The source of that decision separates these products more clearly than whether they run on a schedule.
Notion documents objectives with owners and calculated progress, while win.sh describes an operator that monitors metrics and creates projects from goals. Both can keep knowledge and ask for approval. The choice is whether your team will direct work through its existing workspace or authorize a company operator to route it.
Who does Notion Custom Agents fit?
Notion fits teams that want agents to use the context they already maintain together. Its Custom Agents guide describes schedules, Notion changes and Slack events, with instructions, connected tools and sub-agent delegation. A routing agent or scheduled report can remain beside the pages and databases people already update.
The workspace also supports human-maintained objectives. Its objectives guide documents owners, measurable key results and calculated progress. That is useful when the team wants to keep its existing progress records and add automation around them.
Be precise about the products involved. Personal Notion Agent handles on-demand work, shared Custom Agents handle background jobs, and the separate Claude agents beta adds coding tasks and pull requests. Workers add custom code in beta. These capabilities should not be combined into a claim that every Notion agent runs every kind of job in the same way.
Who does win.sh fit?
win.sh suits a founder who wants a company operator watching business signals and deciding how to route authorized work. Its operating model takes a goal and creates a plan with projects, deadlines and priorities. The operator follows an authority matrix that controls what different kinds of work can do alone.
Its local harness documentation adds editable loops in your own repository, using supported coding agents with your own agent subscription. Loops can select the next useful run time and retain journals. Hosted managed execution is also available, so local operation is an actual option in this comparison.
Confirm the connections your intended job needs. The integration catalog marks Stripe, Plausible and Google Search Console live, while GitHub, custom API and custom MCP are marked soon. Other pages say GitHub is available and enterprise customers can add custom integrations. That conflict requires confirmation for your account before a connector becomes part of the plan.
Where does each product put human control?
Notion's connection controls let individual tools run automatically or always ask. Instructions and access are configurable before you save an agent. For a scheduled customer update, inspect both the source pages it can read and the approval setting on the tool that sends the message.
win.sh applies authority rules by type of work through its company operator. Its published pages do not establish separate action settings for each agent. A founder can therefore evaluate a category such as publishing under the authority matrix, but should not assume it offers the same per-agent control surface as another product.
Memory and authority also deserve separate checks. Notion keeps viewable memories that can reflect corrections, but its public pages do not describe permission-change proposals based on approval history. win.sh says approvals, edits and rejections become operating rules. Its Bug Autofix contract can propose more authority after at least 10 successful fixes and other checks, but the documented upgrade proposal is not tied to approval history.
How repeated work gets verified
Notion's agent testing guidance asks you to define done, run manual tests and inspect outputs and logs. Its public pages do not establish automatic evaluation of delivered results against defined requirements. This matters for a recurring report whose numbers must be checked before it is shared.
win.sh's Bug Autofix contract supplies a more specific verification example. It requests regression tests, defines ordered executor steps and checks error rates against success and failure criteria after deployment. Its documentation does not establish independent verifiers for individual steps or a requirement-level check for every task.
The local harness and hosted platform also have different integration boundaries. win.sh's public API documents snapshot reads and says hosted write endpoints are not public yet. Notion documents hosted MCP connections and Agent APIs for starting chats, submitting actions and reading session history. Specify whether your workflow needs to read a signal or change a connected record before comparing API availability.
Budget for the work you will run
Notion's pricing page lists Business at €19.50 per member per month billed yearly. Purchased credits are $10 per 1,000 monthly or $13 per 1,000 annual credits. Its hosted agents and Notion-hosted Workers do not establish a customer-operated local runtime.
win.sh's pricing page gives self-serve monthly budgets from $50 to $10,000, a hard limit and optional top-ups. Those budgets and Notion's member plans and credits describe different purchases. Use the intended loop or report to estimate what you will fund, without treating the published amounts as comparable execution units.
Notion Custom Agents and win.sh, point by point
The table presents both vendors' answers to the same questions, including sources and individual check dates. Read the explanation for each capability, especially where a local loop or beta changes its scope.
Both columns answer the same questions from each company's public pages. Yes means the company publishes the capability the question asks about, Partly means something narrower, No means it says it does not offer it, and Not published means its public pages do not describe it. Every answer links to its source and shows when we last checked it. The oldest check on this page was November 13, 2026.
| What you compare | Notion Custom Agents | win.sh |
|---|---|---|
| Goals |
Notion Custom Agents
Partly
Notion documents objectives with owners, measurable key results and progress tracking. Its example database calculates progress from current, initial and target values, rather than requiring an agent to own the goal. |
win.sh
Yes
win.sh monitors business metrics, takes a goal and creates a plan with projects, deadlines and priorities. Its company operator routes the work under your authority rules. |
| Leads |
Notion Custom Agents
Partly
Managers can oversee projects and team workloads, and Custom Agents can delegate parts of a job. Its docs do not establish a named lead role responsible for reviewing and coordinating a goal, project or team. |
win.sh
Yes
The company operator reviews metrics, routes work and creates projects from goals. Its pages do not describe distinct goal, project and team leads with their own coordination responsibilities. |
| Tasks |
Notion Custom Agents
Yes
Notion tasks have assignees, dates and status. The separate Claude agents beta accepts assigned tasks and keeps sessions and activity on the page or task where the agent works. |
win.sh
Partly
win.sh lists task delegation, and the local harness records briefs, approvals and outcomes. Its public pages do not describe a shared task record with named human or agent ownership and complete history. |
| Recurring work |
Notion Custom Agents
Yes
Custom Agents run on recurring schedules, Notion changes and Slack events, including messages, mentions and reactions. You can combine trigger types and filter when an agent runs. |
win.sh
Yes
Schedules, events and signals can start a run, and local loops choose their next useful run time. The documentation does not describe a stable root task with a separate task for every occurrence. |
| Workflows |
Notion Custom Agents
Partly
Custom Agents follow instructions and hand work to sub-agents, and Workers add custom code in beta. Its docs do not establish a fixed step-and-branch process for agent runs, and recommend letting the agent determine how to reach the outcome. |
win.sh
Yes
Editable loop contracts specify triggers, authority and verification, and the bug-fix loop defines ordered executor steps, conditions and follow-up. The documentation does not describe independent verifiers for individual steps. |
| Autonomy |
Notion Custom Agents
Yes
Custom Agent connections let you set individual tools to run automatically or always ask. Users receive confirmation prompts before actions, and unexpected external links generate a notification for approval. |
win.sh
Yes
An authority matrix sets what each type of work can do alone or send for approval, and the company operator follows those rules. Current pages do not establish separate action settings for each agent. |
| Learned autonomy |
Notion Custom Agents
Not published
Its docs do not describe proposals to change an agent's permissions based on your approval history. Memories retain context, while approval settings control when the agent asks. |
win.sh
Partly
Approvals, edits and rejections become operating rules, and Bug Autofix can propose more authority after at least 10 successful fixes and other checks. The documented upgrade proposal is not tied to approval history. |
| Memory |
Notion Custom Agents
Yes
Custom Agents can retain viewable memories across chats and update them from corrections. Personal Notion Agent also keeps a Memories database where you can edit or delete saved context. |
win.sh
Yes
win.sh saves decisions, rules and learnings as editable company knowledge, and local loops keep journals. Its pages do not document restoring notes or separate agent, project, goal, team and workspace memory scopes. |
| Planning |
Notion Custom Agents
Yes
Personal Notion Agent can turn a chat request into a launch plan with tasks and owners. Custom Agent creation also lets you review and refine draft instructions, triggers and access before saving. |
win.sh
Yes
Conversational onboarding analyzes a website and creates a 30-day plan, and a goal becomes projects. The pages do not describe reviewing acceptance criteria before each agent task starts. |
| Intake |
Notion Custom Agents
Yes
Task routing agents capture incoming work and route it to a team. Slack messages, mentions and reactions can start a Custom Agent, while connected mail can trigger work from incoming email. |
win.sh
Partly
win.sh describes support issue triage, and its hosted API documents snapshot reads. The API says hosted write endpoints are not public yet, and the cited pages do not establish routing incoming requests to an owner or agent. |
| Connectors |
Notion Custom Agents
Yes
Custom Agents support a catalog of hosted Model Context Protocol connections and custom servers, with scheduled and event-triggered runs. The Agent APIs also let integrations start chats, submit actions and read session history. |
win.sh
Partly
The catalog marks Stripe, Plausible and Google Search Console live, but marks GitHub, custom API and custom MCP soon. Other win.sh pages say GitHub is available and enterprise customers can add custom integrations, so availability is unclear. |
| Models |
Notion Custom Agents
Yes
Auto selects models per task and can switch models during demanding steps of a longer job. It is the default for personal Notion Agent and Custom Agents, subject to workspace model restrictions. |
win.sh
Partly
The public model router recommends a model and fallback using task, quality, speed and price. Its documentation does not establish that hosted workflows automatically choose a model at each step. |
| Cloud or your own computer |
Notion Custom Agents
Partly
Custom Agents run inside Notion, and Workers run custom code on Notion's servers. Its docs do not describe running these agents on your own computer or infrastructure. |
win.sh
Yes
win.sh offers hosted managed execution or an open-source local harness that runs Codex or Claude Code in your own repository. |
| Your own subscription |
Notion Custom Agents
Partly
Business and Enterprise members can connect a ChatGPT Plus, Pro, Business or Go plan to personal Notion Agent for selected GPT models. This does not work with Custom Agents or Auto. |
win.sh
Yes
win.sh's local harness lets you run loops with your own agent subscription using Codex or Claude Code. |
| Pricing |
Notion Custom Agents
Yes
Business is €19.50 per member per month billed yearly on the fetched pricing page. Agent usage draws credits, with purchased Notion credits at $10 per 1,000 monthly or $13 per 1,000 annual credits. |
win.sh
Yes
Self-serve monthly budgets range from $50 to $10,000, with a hard limit, optional top-ups and a seven-day money-back guarantee. |
| Correctness |
Notion Custom Agents
Partly
Notion recommends defining done in instructions, manually testing runs and reviewing their outputs and logs. Its docs do not establish automatic checks of delivered results against defined requirements or evaluation criteria. |
win.sh
Yes
Bug Autofix asks for regression tests and checks the error rate against defined success and failure criteria after deployment. It does not document independent pass, fail or uncertain checks for every task requirement. |
For software work
| What you compare | Notion Custom Agents | win.sh |
|---|---|---|
| Code review rules |
Notion Custom Agents
Partly
The separate Claude agents beta accepts instructions for coding style, branch names and pull request descriptions. Its docs describe teammates reviewing changes, not an automatic code review against repository rules. |
win.sh
Not published
Bug Autofix directs the agent to add a regression test and run checks. The public pages do not describe reviewing changes against discovered repository rules or screening proposed rules against examples. |
| Adversarial review |
Notion Custom Agents
Not published
Its docs do not describe an independent second review of risky code changes. The Claude task-board guide recommends keeping your normal teammate review of the diff and merge. |
win.sh
Not published
Bug Autofix asks for risk notes and tests. The public docs describe no independent two-reviewer adversarial review for high-risk changes. |
| Risk rating |
Notion Custom Agents
Not published
Its docs do not describe a risk rating for a code change before merge. The Claude task-board guide describes flagging risky scope in a product requirements document and leaving merge approval to a teammate. |
win.sh
Partly
The bug-fix executor reports remaining risk and test output with a pull request or patch. It does not document a merge-risk rating for the exact revision. |
| Pull requests |
Notion Custom Agents
Partly
The separate Claude agents beta opens GitHub pull requests linked to Notion tasks and accepts direction through task comments. Its docs do not establish handling review feedback from GitHub pull requests. |
win.sh
Partly
Bug Autofix instructs its local executor to open a pull request or provide a patch, and requires approval before merge. Its public contract does not describe exact-commit approval or addressing review feedback. |
A fact we cannot find or cannot check is left out, and prices can change after the date shown. Task Machine publishes this comparison and is an interested party. Notion Custom Agents and win.sh are named only to identify them, and there is no relationship between the companies. Has something changed? Tell us.
Who each one suits
Pick Notion Custom Agents if shared document context and event-triggered work fit how your team already plans. Its public pages do not describe running those agents on your own computer, and its guidance recommends manual output review.
Pick win.sh if a company operator, adaptive loops and a local coding harness match your operating model. Confirm integration availability first, and treat the Bug Autofix verification contract as a specific loop's behavior.
Where Task Machine fits
Task Machine distributes ownership across people and agents on Tasks, with Goal leads reviewing linked Metrics and proposing missing work for approval. Its Workflow builder adds defined steps, branches, human decisions and verifiers that return failed results. That fits a team wanting explicit shared work records and review boundaries across repeated jobs.
Task Machine is the wrong pick if your priority is keeping all automation attached to Notion context or adopting win.sh's single company operator. The Notion comparison and win.sh comparison explain those boundaries. Take one business metric that changed recently and write down who should propose the response, approve it and check the delivered result.