sf-intelligence vs Hubbl

At a glance

The real difference is the unit of work

Hubbl's product is a scan. You connect an org, it audits packages, automations, fields, code and permissions, and returns a prioritised roadmap with issues ranked by severity. Run it monthly and you get a trend line and a number to put in a slide. That number is comparable to other orgs, because Hubbl processes many orgs and markets benchmarking as a headline reason to buy.

sf-intelligence's product is a question. There is no score, no dashboard and no benchmark. You retrieve your org's metadata once, read-only, into a local vault, and from then on an MCP host — Claude Code, Claude Desktop, Cursor — asks the graph things like "what breaks if I delete this field" or "why can't this profile edit that". Every claim comes back attached to a canonical component ID and a confidence tier, so you can see whether a dependency was declared in metadata, parsed out of Apex, or matched by heuristic pattern.

Those are not competing answers to one question. They are different questions. A scored audit tells you where to look across an org; a cited graph tells you what is actually true about the one component you are about to change.

Pricing and licensing models

Hubbl is priced per org rather than per user. Its pricing page describes a free tier — one scan of one org, covering usage and recommendations across packages, automations, fields, code and permissions — plus a Professional plan with unlimited users, unlimited scans and process views, scheduled scans and ongoing monitoring, and an Enterprise plan that adds API and Hubbl MCP access, SSO, included sandboxes and premier support. List prices for the paid plans are not published; you request a quote. For a large admin team the per-org basis is a genuine advantage: adding twenty more users costs nothing.

sf-intelligence is free and source-available under MIT + Commons Clause — deliberately not "open source", because the Commons Clause withholds the right to sell the software itself as a service. You can read it, run it, fork it for internal use and audit exactly what it does with your metadata. There is no seat count, no org count, no contract and no vendor to notify. See what the licence does and does not permit before you build anything commercial on top of it.

Worked example: one field, two workflows

Say Widget__c.Status__c looks dead and you want it gone. In an audit-shaped tool, the field shows up in an unused-fields section of a scan with a severity and a recommendation — useful for building the cleanup backlog in the first place. In sf-intelligence you interrogate that one field directly, before touching it:

› is it safe to delete Widget__c.Status__c?
▸ verdict: review-first — 3 dependents found
  ValidationRule "Widget_Status_Required" · Flow "Widget_Router" · Layout "Widget Layout"
  cited: CustomField:Widget__c.Status__c
  coverageCaveat: layouts and validation rules fully modeled; dynamic Apex not checked

The two motions compose cleanly. A scan is good at finding candidates across the whole org; a graph is good at clearing a specific candidate. If you run both, use the audit to produce the list and the graph to defend each deletion. The method is written up in full in the dependency-first field cleanup guide.

Choose sf-intelligence when you want to…

• Ask impact, permission and debt questions every day from inside an MCP client, at no cost
• Keep org metadata on the machine — nothing is uploaded, and the vault is a directory you own
• Get a canonical component ID and a confidence tier on every claim, and an explicit coverage caveat when a verdict rests on absence
• Guarantee no write path: the product has no deploy, edit or delete capability to misconfigure
• Run a DIY agent-readiness audit without a contract or a procurement cycle

Choose Hubbl when you want to…

• Benchmark your org's health and complexity against the wider ecosystem — we genuinely cannot do this, and no local single-org tool can
• Discover and monitor business processes, not just metadata structure — a capability sf-intelligence has nothing equivalent to
• Hand leadership a scored, severity-ranked deliverable with a step-by-step remediation roadmap
• Re-scan on a schedule and watch a trend, with no local install and nothing to refresh by hand
• Cover an unlimited number of users on a per-org price, or work partner, ISV and portfolio motions across many customer orgs

When not to use sf-intelligence

Worth saying plainly, because it is the fastest way to avoid wasting your afternoon:

• You need to change something. There is no write path — no deploy, no bulk edit, no rollback. It reads and explains; you make the change yourself in Salesforce or your DevOps tooling.
• You need a shared dashboard. Answers live in whoever's editor session asked. There is no web UI, no scheduled report and no place for a non-technical stakeholder to log in.
• You need comparison against other orgs. The graph knows only the vaults you built. Any "how do we compare to similar companies" question is out of scope by construction.
• You need continuous change capture. The vault is a point-in-time snapshot and goes stale the moment someone edits the org; freshness depends on you re-running a refresh.
• Nobody on the team will use a terminal. It is an MCP server driven from a CLI and an AI host. An admin-only team with no developer is better served by a hosted product.
• You need record data. The vault holds metadata, not business records. "How many open Widgets" is a report, not a graph query.

Try it before you decide — no org required

You do not need a Salesforce org, an authenticated sf CLI or any configuration to see what the answers look like. One command registers a server that ships with a synthetic org — "Verdant Energy", a fictional solar installer — entirely offline:

claude mcp add --transport stdio --scope user sf-intelligence-demo -- npx -y sf-intelligence demo

Ask it a delete-impact or permissions question and judge the citation discipline on fabricated metadata, with nothing of yours at risk. When you are ready for the real thing, connecting your own org is one read-only retrieve.

More comparisons

The closest enterprise peer to Hubbl in this set is Elements.cloud, which also pairs hosted change intelligence with an MCP story. If your interest is a hosted natural-language chat over the org instead of a scored audit, see dx0. Full set: compare hub · dx0 · Metazoa · Salesforce MCP · Sweep.