Product guide.

How the product works

OutcomeCraft turns customer-owned knowledge and recent X conversations into reviewable original posts and replies.

The application is multi-tenant: each organization has isolated users, connected accounts, knowledge, editorial profiles, content and usage records. A user can belong to a team and an organization can connect more than one X account.

↑ On this page

Invitation and first login

An owner invites each person and assigns an initial role.

1. Open the one-time invitation link from your email.
2. Choose Continue with Google and use the exact invited email. This is the recommended path and takes you directly to the workspace after Google consent.
3. If you cannot use Google, expand Use email and password instead, enter your name and create a unique password of at least 12 characters.
4. Open Security, verify your email when applicable, enable an authenticator app and save the recovery codes in your password manager.
5. Review active sessions and remove test or unknown devices.

No invitation yet

Google sign-in never creates an uninvited account. Join the public waitlist; Tecnicora reviews each request and sends invitations gradually. A waitlist request does not create a workspace or guarantee access.

Roles

Role	What it can do
Viewer	Read content, connections, knowledge, publishing and usage information.
Editor	Manage editorial profiles, examples, calibration, knowledge inclusion and content workflows.
Owner	Manage teammates, roles, provider connections, publishing controls and organization settings.

↑ On this page

Portal navigation

Each screen represents one part of the operating workflow.

Overview

Readiness summary and shortcuts.

Daily Content

Current artifact, candidates, tweet of the day and suggested replies.

Performance

Published posts and their latest recorded X metrics.

Editorial Studio

Strategy, voice, examples, calibration and version history.

Usage

Credit balance plus model and X API consumption attributed to your organization.

Connections

Authorize independent X accounts.

AI Connectors

Connect and revoke Claude or ChatGPT operational access.

Knowledge

Review customer-approved capsules sent through your AI assistant.

Publishing

Review the automatic schedule and pause publishing.

Team

Invite, activate and assign roles.

Security

MFA, verification and active sessions.

Use the language selector in the navigation to save English or Spanish (Mexico) as your personal interface preference.

↑ On this page

Connect Claude: step by step

Use your Claude account to prepare approved onboarding material without granting OutcomeCraft access to all of your sources.

Evaluate it before connecting X

Ask Claude: Use OutcomeCraft's evaluate-without-X workflow. Help me complete and calibrate my editorial setup, but do not connect X or publish anything. Claude can configure the workspace with an empty watchlist, build the editorial evidence and generate six explicitly non-publishable calibration samples. Editorial readiness and X readiness are reported separately.

Claude Pro, Max or Free

1. Open claude.ai in a web browser.
2. Go to Customize → Connectors.
3. Select +, then Add custom connector.
4. Name it OutcomeCraft and paste the connector address above.
5. Under OAuth client, choose Use Anthropic's hosted client metadata (Recommended). OutcomeCraft supports Claude's recommended Client ID Metadata Document flow.
6. Select Add, then Connect. Claude opens OutcomeCraft in your browser.
7. Sign in to OutcomeCraft, choose the correct workspace, review the permissions and select Approve connection.

Claude Team or Enterprise

1. An organization Owner or Primary Owner first opens Organization settings → Connectors.
2. Select Add → Custom → Web, enter OutcomeCraft and paste the connector address.
3. Under OAuth client, choose Use Anthropic's hosted client metadata (Recommended).
4. After the owner adds it, each member opens Customize → Connectors, finds OutcomeCraft and selects Connect.
5. Each member signs in to OutcomeCraft and approves access for their own account and workspace.

Enable it in a conversation

1. Start a new Claude conversation on web, desktop or mobile.
2. Select the + button beside the message box, then Connectors.
3. Turn on OutcomeCraft for that specific conversation. Copying a prompt does not turn it on automatically.
4. Only then send: Use OutcomeCraft to show my editorial onboarding status. Do not write anything yet.
5. If Claude returns your workspace status, the connection works.

Configure and operate your editorial playbook

Then send: Help me configure my OutcomeCraft editorial playbook. Interview me first. Use information from this conversation and sources I explicitly choose. Before every write or control change, show the exact proposal and ask for my approval.

Claude can prepare, version and activate profiles; manage examples and approved capsules; run calibration; connect and assign X accounts; configure the watchlist and schedule; generate and read the daily artifact; publish one selected original candidate; administer workspace members; and report status and usage. Provider OAuth still requires opening the returned URL.

Not sure what a profile field means? Ask, for example What does "voice traits" actually change in what gets written? — Claude can read a field-by-field reference and explain it before asking you to fill it in.

Feedback from your AI

After connecting, you can ask Claude to report a bug, request an improvement, find existing requests, read their discussion or support one. Claude searches the shared backlog first and must show you a privacy-safe ticket or comment before requesting approval to write it.

Permissions and revocation

editorial:read reads onboarding and operational configuration. editorial:write permits role-authorized editorial and workspace actions. Every write requires approval of the exact proposal, and sensitive controls—including activation, team changes and publishing—require a separate action-specific confirmation. OutcomeCraft never exposes provider credentials, passwords or MFA material. Suggested replies cannot be published through MCP.

Disconnect it either in Claude → Customize → Connectors or in OutcomeCraft → AI Connectors. OutcomeCraft access tokens expire after 12 hours and can be revoked earlier.

If it does not connect

Symptom	Check
Claude does not show Add custom connector	Confirm you are in Customize → Connectors. On Team/Enterprise, ask an organization Owner to add it first.
Claude shows multiple OAuth client options	Choose Anthropic's hosted client metadata, marked Recommended. Automatic registration (DCR) remains a compatible fallback.
OutcomeCraft says you are not invited	Use the exact email invited to the workspace or ask its Owner for an invitation.
Connector exists but Claude does not use it	Enable OutcomeCraft for the current conversation through + → Connectors.
Authorization expired	Select Connect again and repeat OutcomeCraft approval.

Claude's interface can change. See Anthropic's official remote MCP connector guide.

↑ On this page

Connect X

X authorization is stored per organization and per installation.

Connect an X account

1. Open Connections and choose Add X account.
2. Sign in to the intended X identity and review the requested permissions.
3. Return to the portal and confirm the username shown.
4. Repeat for additional accounts. Each authorization remains independent.
5. Select exactly one default publishing account before enabling any publishing workflow.

↑ On this page

Manage knowledge

Knowledge gives the writer grounded context; it does not override editorial or security rules.

1. Connect the source you prefer to Claude or ChatGPT.
2. Ask your assistant to identify only material relevant to the editorial strategy.
3. Review the exact proposed capsule, its provenance and optional source URL.
4. Approve it explicitly before the assistant sends it to OutcomeCraft.
5. Exclude a capsule later through MCP or the Knowledge screen if it should stop influencing generation.

analyze_knowledge_coverage checks active pillars, knowledge types, provenance, freshness and possible overlap. get_workspace_recommendations combines that result with profile, examples, calibration and X readiness. Duplicate or contradiction flags are review suggestions, never automatic factual decisions.

Good sources

• Positioning and product strategy.
• Approved content pillars and frameworks.
• Public website and published articles.
• Customer-agnostic operating principles.
• Approved writing examples.

Do not include

• Passwords, API keys or connection strings.
• Private customer names, commercial terms or incident details.
• Unapproved drafts presented as facts.
• Material owned by another tenant.

↑ On this page

Teach strategy, tone and voice

The system uses a structured editorial profile and evidence—not a separately trained model.

1. In Editorial Studio, create a draft profile.
2. Define positioning, audience and the business goal of the account.
3. Select English or Mexican Spanish as the editorial output language. For Mexican Spanish, choose neutral, tú or usted audience address and add vocabulary guidance.
4. Add voice traits, content pillars, formats and objectives.
5. Write positive guidance and explicit patterns to avoid.
6. Record prohibited topics, confidentiality boundaries and any additional editorial note.
7. Add several real examples that sound right and several that do not.
8. Generate six calibration posts and mark each on-brand, reject or rewrite with a reason.
9. Activate the profile only when the examples consistently reflect the intended voice and regional language.

Profiles are versioned. MCP-created and web-created drafts appear in the same history. View any version, create a new draft from it, promote a draft for a specific X account, or roll that account back without changing another account's publishing profile.

↑ On this page

Daily content workflow

Treat the recommendation as an editorial decision, not an unattended model output.

1. The scheduled refresh—or an explicitly approved create_daily_content MCP request—reads recent posts from the configured watchlist and account history.
2. The writer uses the assigned profile, approved examples and included knowledge to create distinct candidates and contextual replies.
3. An independent critique scores quality and rejects candidates that fail the thresholds.
4. Open Daily Content in the portal, or read the stored artifact with get_daily_content, including its stable artifact_id, each candidate_id, the selected tweet_of_the_day_candidate_id, gates, critique, source context and suggested replies.
5. Check factual grounding, tone, repetition, length and whether the post adds a distinct point of view.
6. Copy the final text or explicitly approve publish_daily_candidate. Manual publishing works independently of the automatic schedule; replies remain human-review-only.

Final review checklist

• Complete thought and natural language
• No confidential or invented claim
• No unnecessary “AI” opening or generic slogan
• Aligned with an active pillar and objective
• Within OutcomeCraft's 160-character automated-post limit
• Not substantially similar to recent owned posts

↑ On this page

Publishing and safety

Original daily posts and replies have separate controls.

The portal shows whether automatic daily publishing is enabled or paused and the organization schedule. The portal can always pause an existing schedule; enabling production automation is an explicit administrative decision.

1. Confirm the correct default X account.
2. Review multiple days of generated content manually.
3. Confirm the schedule and timezone.
4. Validate that X authorization has the required write permission.
5. Approve automatic original-post publishing only after acceptance testing.

Status meanings

Not attempted

No publishing request has run for that recommendation.

Published

X accepted the post and returned its identifier.

Error

The attempt failed; inspect the safe error, authorization and account selection before retrying.

↑ On this page

Usage and cost

Usage records support customer estimates, internal controls and billing.

The Usage screen presents OutcomeCraft services, the scope of each service, included allowance, fixed-price extra blocks and prepaid balance. For example: Analyzed recent X activity from @username — 10 posts over a 7-day window. The count and time window explain analysis scope; the charge is for OutcomeCraft analysis and processing.

Final Launch Pricing v2 includes up to 700 analyzed watchlist posts per billing cycle on Starter and 1,400 on Pro. Need more? Add 500 analyzed posts for $10. The fixed block price never changes because of provider costs, caching, retries or shared retrieval. The portal warns at configured thresholds before additional processing begins. Watchlist limits are 10 accounts on Starter and 20 on Pro.

Provider operations, model-token economics, retries, shared retrievals and infrastructure costs are kept in a separate internal cost ledger and are not represented as products sold to the customer.

Estimate before scaling

1. Run a representative test period.
2. Review analyzed-post usage.
3. Compare it with the plan allowance.
4. Fund prepaid balance before authorizing a $10 extra block.
5. Re-estimate after changing monitored accounts or publishing frequency.

↑ On this page

Team and security operations

Access should follow least privilege and remain easy to revoke.

• Invite people individually; never share one administrator login.
• Assign Viewer, Editor or Owner according to actual responsibility.
• Keep at least one active owner and remove temporary test members.
• Require verified email and MFA for owners and editors.
• Review active sessions regularly and revoke unfamiliar devices.
• Use recovery codes once, then regenerate or reconfigure MFA if they are exposed.
• Disconnect provider accounts that are no longer used.

Credentials belong in a password manager and runtime secret vault. Never paste them into posts, editorial examples, Notion pages, Slack, source code or support screenshots.

↑ On this page

Troubleshooting

Start with the visible status and avoid reconnecting or rotating everything at once.

Problem	What to check
Cannot sign in	Correct tenant URL, verified email, password, invitation status and MFA clock/code.
X connection fails	OAuth callback URL, client credentials, app permissions and the X account used during consent.
Knowledge is missing	Ask Claude or ChatGPT to list approved capsules, then send the missing source summary only after reviewing it.
Writing sounds generic	Active profile, concrete positive/negative examples, calibration reasons and included knowledge.
No daily recommendation	Scheduler health, tenant timezone, watchlist sources, provider credentials and generation quality decision.
Publishing error	Default account, Read and Write authorization, token freshness, post length and duplicate recovery.
Unexpected usage	Daily breakdown, model, number of refreshes, X operation type and recent configuration changes.

When requesting support, send the organization name, approximate time, screen/status, and safe error text. Never send passwords, tokens, secrets or complete database URLs.

↑ On this page

Waitlist and access

Requesting access and receiving an invitation are separate steps.

1. Open Join the waitlist from the login page.
2. Review the privacy notice, provide minimal qualification information and consent to contact.
3. Submission does not create an account. Tecnicora reviews the request.
4. If accepted, open the one-time invitation and continue with the exact invited Google account or use the password fallback.

Repeated submissions do not improve priority and remain one lifecycle record. Contact support to withdraw or exercise privacy rights.

↑ On this page

Knowledge onboarding options

Choose the smallest access path that produces useful, approved context.

AI connector (recommended)

Connect sources to Claude or ChatGPT, synthesize only what matters, then send approved summaries as capsules.

Manual fallback

Type strategy, boundaries and examples directly in Editorial Studio.

A capsule stores approved text, provenance and an optional source reference. OutcomeCraft does not create a live connection to Drive, Notion or another source. Never place secrets or customer-confidential material in a capsule.

↑ On this page

Billing, data lifecycle and support

Know what is measurable in the portal and what requires a verified request.

Usage values support forecasting; the written offer/order form defines chargeable prices. Revoke sessions, members, X and AI connectors from their portal screens. For full export, correction, restriction, organization deletion, privacy/ARCO requests or billing questions, follow Support, billing and data lifecycle.

↑ On this page

Administrator launch checklist

Complete these controls before onboarding a production customer.

• Custom domain and TLS configured
• Production email delivery tested
• Google and X callback URLs updated
• App-scoped deployment credentials stored safely
• Database backups and rotation procedure verified
• Owner account verified with MFA
• Tenant isolation acceptance tests passed
• Two X accounts tested independently
• Knowledge inclusion reviewed
• Editorial calibration accepted
• Usage metering reviewed against provider records
• Automatic publishing remains paused until written approval

↑ On this page