Sign In

The screen has four parts: a top bar (your account on the right), a left rail of sections, the map filling the middle, and a detail panel that slides in whenever you open a nation.

The left rail is grouped by how you work. Pipeline is the flow of new signal into action: News → Opportunities → Projects (administrators also see Research here). Library is the reference shelf: Contacts, Directory, Settlements and Legal. Issues, Guide and (for administrators) Admin sit at the bottom.

The Directory’s search box jumps straight to any of the 638 First Nations by name or band number — click a result and the map flies there.

The map shows every First Nation as a pin. Click a pin — or search by name — and the map flies to that nation and opens its card.

The Settlements pane drives what the map shows: pins shade red → green by total federal share (amber = a claim in progress), and the fiscal-year picker spotlights a single year’s settlements. Below it, Who received what covers all 638 nations: in All years the recipients rank by dollars — every line reads amount · count — and the nations with no settlements follow under their own header at $0 · 0 in red, matching the map’s red-means-none. A fiscal year shows that year’s recipients only. Tap any amount for the per-claim breakdown; the CIRNAC source and its as-of date are printed right on the pane. About The Data explains the figures.

The Directory is the same set of nations as a searchable list, for when you would rather scan names than hunt on the map. Rows with settled funds show the amount right in the line, so the list scans for capital at a glance. Clicking a row opens the nation exactly like clicking a pin.

You can flag a nation green or red with a short note — a quick signal for the team (“strong relationship here,” “handle with care”). Flag notes also feed the opportunity engine, so they are worth keeping current.

Every number in the portal has a source, and knowing each source’s rhythm is the key to trusting what you read:

• Settlements & claims — CIRNAC, refreshed yearly. The dollar figures come from Crown-Indigenous Relations’ own Settlement Report — the official record, including exact amounts that often never appear in the news. CIRNAC publishes each summer with figures “as of” June 30, and the portal’s database is refreshed from it; the as-of date is shown on the data itself.
• News — automatic, around the clock. The feed pulls its sources every 15–30 minutes, day and night. A settlement announced this morning is in the feed within the hour — and can raise an opportunity alert.
• How the two meet — News knows it happened; the settlements table holds the official figure. Anything settled after the current as-of date shows up in News right away and gets its official amount at the next CIRNAC refresh.
• Leadership — the ISC First Nation Profiles registry, fetched on demand and cached with its date. Histories come from Wikipedia the same way. Legal is curated from primary sources (Supreme Court and Specific Claims Tribunal decisions) on a standing review.

Opening a nation brings up the detail panel: a stack of foldable sections you can read top to bottom. Click any section header to collapse or expand it, or use Collapse All / Expand All at the top.

• The card opens on two things — Claims status (the money, with evenly sized buttons that open a window of organized claim data: status, federal share, the full date timeline from filing to conclusion, the description, and the CIRNAC sources) and Leadership (who to call). Every other block starts folded — each folded header still shows its count, total, or status, and whatever you open or close stays that way.
• Claims & settlements — the nation’s Specific Claims position and settlement history, each dated, every money row carrying its own source link to the deepest public record.
• Leadership — the current Chief & Council from the official ISC registry, with band-office contact details. Former leaders archive into a “Past leadership” fold rather than disappearing.
• Needs & capacity — amber need chips and green capacity chips you add by hand, from a fixed vocabulary so matching stays clean.
• Flags — any green/red flag on the nation and its note.
• Relationship standing — the strongest stage your team has reached with any contact at that nation.
• Recent news — headlines matched to this nation.
• Opportunities & potential partners — suggested routes for this nation, and nearby nations whose strengths could answer its needs. Each suggestion carries its own ＋ Project, so a matched pathway becomes a working project in one click — evidence, funding record and the legal precedent linked in.

Every block that comes from an outside source shows where it came from and when (“ISC Registry,” “as of <date>”), so you always know how fresh a figure is.

To keep cards fast, some data is fetched only when you ask for it, then cached so the whole team sees the same result. Look for these buttons inside a nation’s card:

• Pull History from Wikipedia — fetches and stores the nation’s overview.
• Refresh From Registry — pulls the latest Chief & Council and office details from the ISC First Nation Profiles registry. (Administrator action.)
• Scan News — runs a targeted, AI-powered web search for that one nation and files what it finds into the feed. (Administrator action; each scan uses a paid search.)

Contacts is your living relationship record. Add people by hand with ＋ Add Contact, or Import a vCard (.vcf), CSV (.csv) or contacts export (.json).

• Link to nations & roles — connect a contact to one or more nations, each with a role (Chief, Council, Band Manager/Admin, Elder, Band Member, Vendor/Resource…) and a relationship status.
• Relationship meter — five stages describing your team’s engagement with that person: New → Contacted → In dialogue → Working together → Trusted partner. It always describes the relationship — never a rating of the person.
• Attachments — keep files with a contact.
• Comment thread — a running conversation on each contact. Attach a reminder to a comment, or tag a teammate to pull them in.

Opening Contacts shows everyone by default — there is no filter to clear. Use the search box to narrow the list.

News is pulled automatically around the clock from Google News, APTN and GDELT — plus Grok · X when it is enabled — and each item is matched to the nations it affects.

• Why it’s here — a short line under each headline tells you why the item is in the feed: whether a First Nation was named in the coverage, it came from a targeted scan, or it is federal- or provincial-scope news.
• Nation chips — show which nation(s) the item is tied to. Click a chip to filter the feed to that nation.
• AI summaries — once a headline has been summarized, the feed shows a plain-language summary (marked with a teal AI chip) that leads with what matters; the original headline is one hover away.

Opportunities has two tabs — one where you ask, one where the engine tells you:

• Screen — query all 638 nations by funding, need and relationship, ranked by fit. Click a row and the card opens on The Opportunity — a labelled fact sheet: Need (with its status — open, being addressed, or gone from scoring once resolved), Funding (an honest 24-month window), Legal only when a precedent genuinely applies, and Standing. Admins get an AI Backcheck button on the card — one click sends Opus to the live web to re-verify each open need, quoting its evidence with a source link; a need found being addressed is written immediately (the score drops in the same click), while a resolved finding waits for your one-click Mark Resolved, because resolving removes the opportunity from scoring entirely. A click on any need chip still opens Update The Record to set status by hand with a line of evidence. When the AI opportunity screen runs, open needs older than a year are also checked against the nation’s own coverage; with evidence, the AI marks them being addressed (never resolved — that stays a human call). Whatever a backcheck establishes travels into Research with the nation, so paid searches are never bought twice. Being-addressed needs score lower; resolved needs score zero. The score is transparent: every factor that produced it sits in the line under the name. Rows you haven’t reviewed carry a green wash and a New pill under the score — click one and it clears for you, returning only if the score changes. Save Screen keeps a query you like as a one-click chip for next time.
• Alerts — the engine’s shortlist of moments to act. It cross-references funded nations, tagged needs, flags and live news, and only raises an alert on real evidence of an unmet need. Tiers, in order: URGENT (time-critical, or several signals at once) → AI POTENTIAL (a funded nation with a real need surfacing in the news) → FUNDED → POTENTIAL. Triage each alert to Pursuing / Watching / Passed (the view picker filters by that), tune what gets raised with Manage Triggers, and check the engine’s hit rate with Performance.

Good-news stories are deliberately held back — a celebration is not a need. Every opportunity is one click from a live project.

A Project is where the work happens. Start one from an opportunity, or from scratch with ＋ Add Project. The pane has two tabs: Projects (grouped by nation — one button toggles the sort between By Nation and By Recency; click a nation to fold its projects) and My Tasks (everything you’ve been tagged on, across projects and contacts — check one off to move it to Completed).

• Recommend Path — the project AI reads the nation’s funding, needs, news and your contacts, then proposes selectable goals and a step-by-step checklist.
• One conversation (administrators) — the project chat is a research conversation: plan, run the research and drop files right there, or open the full thread in Research. Push a research session into a project and that thread becomes the project’s chat.
• Work it — pick the goals that fit, check off steps (drag to reorder, click to edit), and keep notes as you go.
• Your people, right there — contacts at that nation are suggested inside the project, and a linked contact’s name opens their card without leaving the project.
• Cost, to the cent — every AI action shows its cost in a per-project spend ledger.
• Nothing is lost — Delete moves a project to Archived, where Restore brings it straight back; only Delete Forever, pressed on an already-archived project, removes it for good (administrator action). Its AI-spend history stays on the ledger either way.

My Tasks collects anything a teammate tags to you in a comment, so nothing slips through.

Legal is a curated shelf of Specific Claims and Aboriginal-law precedents. Each entry carries a full citation, a plain-language holding, a practical “Why it worked,” and a link to the source.

Any card with a source is click-to-source — open the ruling in a new tab. This is reference material, not legal advice; that line sits at the top of the shelf and never comes off.

The portal is built so a signal never dies in a feed — you can attach it to the work. There are two ways to link:

1. The chainlink icon on any news row, opportunity, or Legal category opens a “Link to…” box: choose a project or a contact, and the item is attached to it.
2. Inside a project or contact, the Links section groups Linked Contacts/Projects, Opportunities, News, Legal and (on projects) Research — each with a ＋ search picker to attach items directly. A Research chip reopens the live session so the conversation can continue.

Once linked, the item appears in that project’s (or contact’s) Links section, so the whole story travels with the work: the article that started it, the opportunity behind it, the precedent that supports it, and the people involved.

• Reminders — set a reminder on a contact, a project, or a specific comment. When its time passes, the portal emails the recipients. (Sending requires the email service to be configured — see the Admin guide.)
• Comments & tags — the thread on contacts and projects is where the team talks. Tag a teammate and the item lands in their My Tasks.

Found a bug, or want a change? Issues in the left rail lets anyone file an issue or a request, attach a screenshot, and track it. Administrators triage each one with a priority and a notes thread — each new filing emails the administrators, and the Issues icon hues red with a count while anything is open.

Your email, account group and role show in the top-right, with Sign Out beside them. The group is simply whose account it is — Internal (RMFC staff) or Client (licensee firms); both see the full working portal. Everyone also has an Account entry in the left rail (admins see it as Admin) showing their own details and a Change My Password button. Forgot your password? Use Forgot password? on the sign-in screen to get a reset link by email — or ask an administrator to send you an invite or reset link. Administrators can change their own password in the Admin panel (Users & Access → Your account). If a newer build of the portal goes live while your tab is open, a lime bar appears across the top — one click on Reload brings you current.

Everything a team member can — the map, directory, contacts, news, opportunities, projects, legal, links and reminders — plus:

• Create and manage users.
• Run the paid AI sweeps and per-nation news scans.
• Run Research and save it into projects.
• Refresh leadership from the ISC registry.
• Delete saved records — deletes are admin-only everywhere and enforced on the server.
• Review the full activity log.

The Admin panel opens from the left rail (visible only to admins) and has three tabs: Users & Access, AI and Activity Log.

• Create User — fill first / last / phone / email / password / role / group and Create User. They can sign in immediately.
• Send Invite — instead of setting a password, email the person a set-password link and they choose their own. The invite carries the group and names it in the email.
• Roles — User (create, edit, run the free features; no deletes) or Admin (everything).
• Groups — every account is Internal (RMFC staff) or Client (licensee-firm accounts). Both see the full working portal and everyone works from the same shared data. The difference is the machinery: only Internal admins see the AI tab, the Activity Log, Issues triage, and the full user list. Client admins manage Client accounts only — they can add, invite, edit and reset their own people, and every account they create is a Client account. Internal admins can move an account between groups from the Group column; it takes effect at the person’s next sign-in or page load.
• Bell · Edit · Send Reset · Delete on each row. The bell opens that person’s email notifications — three switches (Reminders, Tagged on a note, Issue reports) for which portal emails reach them; invite and password-reset emails always send. Send Reset is one button with two jobs: an invited account gets a fresh invite (its old links stop working), an active one gets a time-limited reset link — emailed, or copyable if email is not yet configured. You cannot delete your own row.

Users can also reset their own password from the sign-in screen. For privacy, that request never reveals whether an account exists — the same neutral message shows either way.

There are two kinds of retrieval: automatic (free, always on) and on-demand (paid, and gated behind the admin panel).

A scheduled process pulls Google News, APTN and GDELT every half hour, matches each article to the nations it names, and files it. This runs with no clicks and costs nothing beyond hosting.

Settlement and claim figures are imported from CIRNAC’s annual Settlement Report (published each summer, figures as of June 30) — the official record, with exact amounts the news often never carries. The refresh is a standing request to the developer: when the new report publishes, say the word and the data lands straight in the live database — no deploy, no downtime; the map layers, year picker and totals recompute at the next sign-in. The Legal library works the same way on a quarterly case law check. Between refreshes, brand-new settlements arrive through News immediately and get their official figure at the next import.

When new articles land, Claude can rewrite each headline into a short, plain-language summary that leads with what matters, and screen the feed for real opportunities. The screen is deliberately strict: Claude must quote the unmet need from the article itself before anything is called an opportunity, and good-news stories are dropped. That is why the Opportunities panel stays signal, not noise.

The AI tab’s sweeps and the per-nation Scan News button send live web searches through Claude and/or Grok, then run the results through the same match-and-screen pipeline. These cost money per search, so they only ever run on an explicit admin click — never on a timer. The drill-in’s AI Backcheck works the same way: a few cents of live search per nation, billed to the same key, itemized in the same ledger.

Pull History (from Wikipedia) and Refresh From Registry (ISC leadership) fetch and cache a single nation’s data when you ask. The project AI — Recommend Path and the chat thread — uses Claude to turn a nation’s funding, needs, news and contacts into goals, a checklist and strategy, and records every call’s cost on the project.

The AI tab has three action cards. Each shows a live cost estimate before you run it and a progress bar while it runs:

• News Refresh — pull every news source (the free feeds plus Claude and Grok web search) and summarize new headlines.
• Legal Refresh — a Claude + Grok web-search sweep for recent court decisions and Aboriginal-law rulings, summarized.
• Opportunities Refresh — refresh News (and Legal), then have Claude review everything and surface real opportunities.

The APIs tab shows every connected outside service — Anthropic (Claude), the optional Anthropic Admin billing key, xAI (Grok), Resend (email) and the news-poker link — each with an Active / Not-configured pill, what it powers, a link to the provider’s key console, and a link to where the secret itself lives. Secrets are baked in at deploy: after changing one, redeploy the portal.

Below the cards sits the model manager: the Claude lineup with its prices, a Default pill on the engine every AI feature follows, and Check For New Models to pull whatever Anthropic has shipped that isn’t configured yet — add one by typing its two prices once and flip it to Default when ready. Then provider cards, and a Claude spend-by-month table that is the complete portal ledger — screening, refreshes, sweeps with their search fees, the project copilot, and all of research including failed runs. If the optional ANTHROPIC_ADMIN_KEY secret is set, an Anthropic account — actual billed block shows the account’s real month-to-date bill straight from Anthropic’s Cost API — the same number as the Console, including Workbench use the portal itself can never meter.

Research is a conversation that ends in evidence. State a goal and you get a runnable plan on the first reply — assumptions folded in, at most one clarifying question ever. Talk it over as long as you like: the plan updates when the direction changes, and a free local scan pins whatever the portal already holds (news, opportunities, settlement digests) onto the thread.

• AI Target — the lens button at the bottom of the left rail arms targeted mode: click any nation, news item, opportunity, legal case or contact and a research session opens anchored to that exact item (Esc cancels).
• Free first — Present Local Data compiles the portal’s own matches into a $0 brief without spending the dive.
• Run when you say so — click Run The Research or just type “run the research.” Nothing paid ever starts on its own; while it runs (and while it plans) a live status line shows the stage and the clock, and real cost shows after the run.
• Findings you can act on — results lead with The Answer — the direct reply to your goal, full roster included when you asked for a list — then the Opportunities — the plays, each with its Next step and its supporting record rows nested right beneath it — and The Rest Of The Record (everything else, who/what/when, sources attached). Click a nation’s name to fly the map to its card. Every opportunity and record card carries Research (a targeted session on that item — the ← in the header brings you back) and ＋ Project, pinned bottom-right; ＋ Turn This Into A Project under the brief turns the whole session into one project in a click.
• Keep digging — the brief’s suggested follow-ups are click-to-insert, and the results chat can run fresh web searches (a few cents, counted on the session) when the record can’t answer.
• Attach evidence — drag a PDF, image or text file onto the thread (or use the paperclip). It is analyzed once against your goal (cents, counted on the session), stored with the session, and its extracted facts ride along in every later turn, the dive and the results chat. Word/Excel: save as PDF first.
• It becomes work — ＋ Add Project at any stage writes the whole conversation into a project and links the live session (the project’s Links → Research) so it can be reopened and continued later. Save As Agent turns a good hunt into a re-runnable agent.
• Standing Rules — after each run the portal distils durable preferences from your answers (verification standards, scope habits) into Standing Rules, reviewable and deletable from the button on the Research pane — every future run follows them.
• Two ends of one rope — every project’s chat is itself a research session (created the first time you use it, at no cost), the project’s Attachments section holds that session’s analyzed files, and linking any research session to a project makes that thread the project chat. Research and projects are the same conversation.

Three outside services do work the portal cannot do alone. The portal only stores the keys; the accounts and billing live with whoever owns them. Each is optional — leave a key unset and that feature simply reports “not configured” and is skipped, while the rest of the portal keeps working.

The cost per AI call has been tuned down, and every paid AI action is gated behind the Admin panel, so spend is predictable rather than open-ended.

Watch the Claude spend-by-month table in the AI tab to see real usage, and adjust your caps from there if the team runs the sweeps heavily.

If RMFC wants to keep the AI and email features under its own control and billing, set up your own accounts and drop in your own keys.

1. Email (Resend or equivalent) — sign up and verify a domain you own (for example, your own RMFC domain). That is what lets reminder and reset emails send from your address. Store it as RESEND_API_KEY.
2. Grok (xAI) — create an xAI account, generate an API key, and store it as XAI_API_KEY. Set your ~$20/month cap on the xAI console.
3. Claude (Anthropic) — create an Anthropic account, generate an API key, and store it as ANTHROPIC_API_KEY. Set your ~$20/month cap on the Anthropic console.

The Activity Log tab is the audit trail — who did what, and when, across the portal. Filter by user and refresh. Meaningful actions (sign-ins, edits, deletes, AI runs) are all recorded.

• Deleting any saved record (contacts, attachments, triggers, alerts, tags, flags, leadership, projects, goals/tasks, users) is admin-only and enforced on the server. Team members may create, edit, and reversibly unlink — but not delete.
• Refresh From Registry (leadership) and Scan News (per nation) are admin actions.
• The paid AI sweeps are admin-only.
• Delete is always the red button; reversible actions (unlink, remove) stay quiet ghost buttons.

• Sign-in is required for everything; sessions last 30 days.
• Passwords are stored only as salted hashes — never in plain text.
• The 638-nation dataset is built into the app; the database holds your team’s work — contacts, projects, tags, flags, news matches and the audit log.
• File attachments live in Cloudflare R2; deleting one is a deliberate admin action.
• Keys for the outside services are stored as Cloudflare secrets, not in the code.