SourceLaunch Trevra
Open-source operating system for agent-run GTM

Spawn the GTM team your company needs.

Give Claude or Codex a revenue outcome. Trevra composes modular skills to research, qualify, prepare outreach, follow up, invoice, and collect—while evidence, approvals, and every run stay under your control.

Launch hosted workspaceExplore the modules
✓ Operate from Claude Code or Codex✓ Self-host on PostgreSQL✓ Update modules through GitHub
trevra / founder-led-growthlive
$spawn gtm --goal "book qualified founder calls"
◉Scoutmarket researchrunning
◉Qualifierfit scoringrunning
◉Operatormessage preparationready
runmodulestate
run_8f2agtm.visibility-auditrecorded
run_8f2bgtm.score-leadrecorded
run_8f2cgtm.outreach-draftprepared
run_8f2egtm.outreach-sendawaiting you
◆Approval required5 personalized emails · send delay 30 min
sha256:4f21b8…c7e0
Typed modulesEvidence on every runPayload hashingWorkspace approval policyOpen-source and self-hostable
One revenue loop

Agents do the work. Trevra makes the work governable.

Use Claude Code, Codex, or another capable agent as the operator. Trevra supplies the contracts, runtime, ledger, evidence, and approval boundary.

01

Give Claude the outcome

Describe the market, offer, constraints, and target result.

02

Trevra composes the run

The agent selects typed modules and records the plan.

03

Safe work runs automatically

Research, scoring, critique, and preparation can run immediately.

04

Consequential work stops

External writes wait behind your workspace policy.

Community module registry

Install the judgement. Fork the parts you disagree with.

Every module declares its inputs, outputs, side-effect class, approval requirement, and version. The public catalog is generated from the executable registry.

8modules shipped in the current registry
GitHub → validation → catalog → hosted runtime
gtm.visibility-auditv1.0.0

AI-visibility audit

Probe a domain for how readable it is to AI shopping assistants and answer engines.

network-readsafe to compose
gtm.score-leadv1.0.0

Score lead fit

Deterministic fit scoring with a human-readable reason for every contribution.

nonesafe to compose
gtm.outreach-draftv1.0.0

Draft outreach email

Produce a deterministic, compliance-footed outreach draft for a lead.

noneapproval declared
gtm.channel-preparev1.0.0

Prepare a channel post

Shape a draft to one channel and run the result through the copy critic.

nonesafe to compose
One contract, any implementationTypeScript today; HTTP and JSON Schema keep the ecosystem polyglot.
Open catalog JSON →
Control plane, not black box

Give agents capability without giving away control.

Models interpret commercial context. Deterministic software controls permissions, state transitions, money, approvals, and external writes.

Complete run ledgerInputs, outputs, evidence, failures, and outcomes are written per workspace.
Cryptographically pinned approvalsThe exact approved payload is hashed before execution.
Scoped delegationLimit action type, confidence, amount, volume, and delay.
◆workspace.policy.yamlenforced
external_writes:
  default: require_approval

outreach.send:
  max_recipients: 5
  confidence_min: 0.84
  delay_minutes: 30

scope.change:
  delegation: forbidden
✓ Evaluated before every consequential run
Run it your way

Hosted convenience. Self-hosted ownership. The same modules.

The public site and catalog deploy at Cloudflare’s edge. The product runtime stays portable: managed by Trevra or deployed into infrastructure you control.

Self-hosted

Own the runtime and every record.

Fork the repository, deploy Express and PostgreSQL, connect your own providers, and keep the ledger inside your environment.

  • ✓ Open code and forward-only migrations
  • ✓ Exportable workspace ledger and evidence
  • ✓ Operator-controlled module configuration
View self-host modules
Questions founders ask

Clear boundaries before you delegate revenue work.

Hosted and self-hosted deployments use the same module contracts and approval model.

How do I actually use Trevra?

You give Claude Code, Codex, or another agent a GTM outcome. The agent calls Trevra modules, while Trevra records the run and stops at the approval boundary for consequential work.

Can an agent send messages or create invoices by itself?

Only inside the policy you set. Consequential actions can require manual approval, and the exact approved payload is hashed before execution.

How are community modules shared safely?

Modules are versioned through GitHub and declare their schemas, side effects, and approval requirements. Validation and tests run before a catalog release.

Can I self-host Trevra and keep my data?

Yes. The application uses PostgreSQL, the runner is open, and the ledger, evidence, configuration, and exports can remain in infrastructure you control.

Start with one loop

Stop being the bottleneck in your own GTM system.

Connect a workspace, install the first modules, and let your agent prepare the next revenue action—with receipts.

Launch TrevraTalk to the founder