The strategic intent layer for AI agents

Your agents ship fast. Do they ship the right thing?

Built for software companies from seed to Series B with 5 to 100 engineers shipping 10+ agent PRs a day. Helen's Foundry gives every agent the live strategic picture, and checks that each PR is tied to a ratified decision before it merges.

Join the waitlist See how it works

The chain behind one ticket

The problem in practice

Your agent followed the ticket. The ticket followed an obsolete decision.

The implementation can be correct while the premise behind it is stale. Code review rarely includes the current decision chain.

One stale-premise scenario

  1. 01 · Monday

    The decision changes

    The team ratifies an enterprise-first strategy.

  2. 02 · Wednesday

    The ticket stays active

    An existing ticket still targets self-service growth.

  3. 03 · Thursday

    The PR looks right

    An agent follows the ticket. The implementation is clean, and every test passes.

  4. 04 · Daily sweep

    Helen raises a finding

    Helen flags the ticket: it contradicts the strategy the team ratified on Monday.

The problem is not bad code. The ticket is clear, but the premise behind it is no longer current.

Who it's for

Built for agent-heavy software companies

Helen's Foundry is in early access for software companies from seed to Series B with 5 to 100 engineers shipping 10+ agent PRs a day.

seed to Series B software companies

Your 5 to 100 engineers ship with agents every day.

Your agents produce enough code that context and review no longer travel at the same speed. Helen's Foundry connects each task and PR to the current decisions above it, so the team catches drift before it compounds.

  • 10+ agent PRs a day across the engineering team
  • Claude, Codex, Cursor, or another MCP-capable workflow
  • Strategy and product decisions that change faster than implementation context

CTOs & VPs of Engineering

Agent throughput has become an executive control problem.

Helen's Foundry lets the engineering leader increase agent throughput without making senior reviewers reconstruct the premise behind every change. The GitHub check shows whether the work serves a ratified decision, not only whether the code passes.

  • Senior review load rising with agent-generated code volume
  • Rework caused by stale, duplicated, or unsupported premises
  • AI spend growing without a defensible measure of aligned output

What changes in your workflow

Keep your tools. Make the decision chain explicit.

Helen works with your existing documents, tickets, repositories, and coding agents. Start with one repository and one product area.

  1. 01

    Draft the decision chain

    Your agent uses existing documents, issues, repository context, and work history to draft the initial context chain.

  2. 02

    Ratify material premises

    An authorized person ratifies the current strategies, goals, and constraints.

  3. 03

    Give agents linked work

    Each agent receives the ticket, its current decision chain, open findings, and specifications.

  4. 04

    Check the result

    The GitHub check shows which decisions each PR serves and holds it when one is not ratified. Findings go to Reconcile, where a human chooses an action.

What your team adopts

Ratify material premises and connect execution to them. Keep your tracker and start with one workstream.

What you get

Catch it when it breaks

The chain is only half of it. Helen's Foundry watches your work against that chain, from strategy down to the pull request, and flags work that no longer fits.

01 · After every change

Drift, surfaced as a finding

Helen's Foundry checks the chain after changes and in a daily sweep. It flags work that no longer fits: a goal that contradicts the strategy above it, a goal with no work built beneath it, or a plan that has not been updated while its work changes. You review each finding and decide what matters.

Contradicts a related artifact Majorseverity
6d ago
Go all-in on enterprise
contradicts
Double self-serve signups
Severity: major.
Helen

Helen:This strategy commits to enterprise-only, while the goal beneath it targets self-serve growth. The two pull in opposite directions. (82% confidence)

02 · On the pull request

The decision chain, where you review

Every check lands as a sticky comment on the PR: which artifacts the change serves, the specifications that apply to it, and any premise that still needs ratification. Your reviewers see the decisions behind the change, inline.

agent-helen bot commented just now

Linked to ticket Prorate seat additions on annual plans

SpecificationSource
Adding seats on an annual plan charges for the rest of the term.ticket Prorate seat additions on annual plans
Invoices show seat changes as their own line.project Seat-based billing

Ratify Usage pricing for mid-market to clear the neutral check.

Coherence checks

24 ways work drifts from intent

Helen's Foundry ships a growing catalog of strategic-coherence checks, from plain structural rules to AI judgment and embedding geometry. Each detector page cites the research it draws on. Every check raises a finding you can reconcile.

Pure rules over the context chain: owners, links, priorities, lifecycle fields.

  • Stale Premise

    Active work still builds on an artifact above it that was abandoned, marked obsolete, or marked a duplicate.

  • Stranded By Closure

    Open work sits under a parent that completed or migrated, so the artifact needs to close or move to an active parent.

  • Coordination Collision

    Two in-flight artifacts under the same parent have different owners working on them at the same time.

  • Lifecycle Contradiction

    An artifact's status fields disagree with each other, such as a closed artifact with no closure reason or an open artifact with a closure reason.

  • Duplicate Singleton

    More than one open, unarchived artifact exists for a type that allows only one in the workspace, such as mission, vision, or strategy.

  • Coverage Gap

    An active persona or problem has no active project, requirement, or story linked to it.

  • Execution Context Gap

    Active stories in a project or requirement are not linked to both an active persona and an active problem. Link the shared project or requirement so its stories inherit the context.

  • Unowned Work In Flight

    Work that is in progress or in review has no owner assigned.

  • No Prime Goal

    The workspace has several top-level goals or strategies with no shared parent above them, so work below them can point in competing directions.

  • Decomposition Gap

    A mission, vision, strategy, goal, or domain model has children, but no execution artifacts beneath it.

  • Priority Vacuum

    Work that is in progress or in review has no priority set, so you cannot rank it against other in-flight work.

  • Missing Accountable Owner

    A committed strategy or goal has no owner accountable for its outcome.

  • Resource Misallocation

    A goal or strategy has urgent work beneath it, but no work under it has an owner assigned.

Built for AI agents

Your agents, with the full picture

Start onboarding in Claude, Codex, or another MCP client. Ask it to use the repository, issues, PR history, existing documents, and working-session context it can already access to draft the initial context chain through Helen's MCP tools. Your team ratifies the material decisions; every connected agent then loads the live context for its work instead of reconstructing it.

https://hf.dev/api/mcp/w/{workspace-id} 30 tools
  • activity_feed_list
  • artifacts_search
  • workspace_external_references_list
  • artifacts_history
  • external_references_list
  • artifacts_create
  • artifacts_set_parent
  • links_add
  • pending_actions_create
  • signal_members_list
  • artifact_types
  • members_list
  • workspace_summary
  • chain_get
  • pr_report_get
  • artifacts_delete
  • artifacts_suggest_parents
  • links_list
  • pending_actions_list
  • signals_list
  • artifacts_list
  • workspace_audit
  • artifacts_get
  • docs_get
  • workspace_external_reference_get
  • artifacts_update
  • demand_context_get
  • links_remove
  • reconcile
  • send_feedback

Shipping with Claude Code? Connecting is one command:

claude mcp add --transport http helens-foundry https://hf.dev/api/mcp/w/{workspace-id}

Any MCP client can connect. Your workspace's exact URL is on Settings > MCP clients once you're in.

Plugs into the tools you already use

Helen's Foundry speaks MCP and ships first-party connectors, so your agents and your team reach the same workspace context from wherever they already work.

  • GitHub
  • Anthropic
  • Linear
  • Notion
  • Figma
  • Slack
  • Discord
  • Sentry
  • Any MCP client

An editor for your thinking, not your prose

Helen is the AI built into Helen's Foundry. She doesn't improve your writing, she interrogates your reasoning.

Helen

This strategy commits to enterprise-only, but the goal beneath it doubles self-serve signups. These two conflict.

  • She shows you the conflict instead of rewriting your work. She checks your thinking, not your grammar.

  • She reads the whole context chain, so she catches when one artifact quietly contradicts another.

  • Every finding she raises is logged, so the reasoning behind a decision outlives the decision.

The team building Helen's Foundry

A two-person founding team shipping the strategic intent layer for AI agents.

Get early access

Free for now. Rough around the edges.

Helen's Foundry is invite-only early access. It's free for early design partners for a limited time, it's rough, and your feedback shapes what ships next. Join the waitlist and we'll set up your workspace and send you a link.

Join the waitlist

Prefer email? Reach us at hello@helensfoundry.com directly.

Free
For early design partners, for a limited time.
Rough
Real value, but expect rough edges.
Yours
Your feedback shapes what ships next.