Trust at the speed of AI.
The programming language and accountability infrastructure for agent-driven development.
KUJO
Software production is outpacing verification.
Agents can report completion. Teams still need a reviewable chain from requested work to release decision.
- 01
The claim
“Tests passed. Ready to merge.” An illustrative agent statement.
- 02
The evidence
Command, exit code, changed revision, environment and retained failure output.
- 03
The gap
Useful logs already exist. Connecting them across tools and review boundaries remains work.
Greater autonomy requires stronger accountability.
The cost of an unverified action grows as agents gain authority.
- 01
Suggest code
Human selects the change → diff review
- 02
Complete a task
Agent edits and runs commands → execution records
- 03
Operate a workflow
Retries and external effects → boundaries, recovery and approval
AI use is broad. Trust remains conditional.
Adoption of AI assistance is not evidence of autonomous production deployment.
- use coding assistants or coding agents at work
- 65.9%
- use AI agents or automated workflows at work
- 26.2%
- trust AI when answers are easy to validate
- 48%
Build with clarity. Operate with control.
The language is the long-term foundation. Existing tools offer an earlier route into engineering workflows.
- 01
Clarity
Kujo language / Spec / Eval: explicit intent and testable outcomes.
- 02
Context
MCP / CaseFile / RunLedger: useful inputs, debugging evidence and run records.
- 03
Control
Dispatch / Workcell / ShipCheck: orchestration, bounded execution and release checks.
Existing repositories + coding agents explicit component contracts Kujo runtime + CLI
Inspect the execution boundary
Workcell owns supported container bounds. Dispatch owns workflow state. Neither confers enterprise certification.
Inspect the evidence boundary
CaseFile captures commands. Eval runs configured checks. RunLedger stores supplied outcomes. Artifacts remain separate.
Accountability spans the whole task.
Each component owns a specific responsibility. A reviewable workflow composes their artifacts explicitly.
- 01
Instruction → Context
Spec / PackWrite → MCP and repository context
- 02
Execution → Evaluation
Dispatch / Workcell → Eval result artifacts
- 03
Evidence → Review → Release
CaseFile / RunLedger → reviewer → ShipCheck gate
Human authority remains explicit at the release decision.
The foundation is already being built.
Inspect the implementation and its limits, then reproduce the smallest useful proof.
- 01
Language and VM
Kujo 1.9.0 is documented; a local 1.9.0 binary executes the recorded fixture tooling.
- 02
Bounded execution
Workcell 1.3 documents Docker/Podman limits, explicit exports and receipt verification.
- 03
Reviewable artifact
This presentation uses the native Kujo builder. Its demo retains a failing command, correction and evaluation.
A claim becomes inspectable evidence.
Recorded local coding fixture: correct a post-money ownership calculation. Inspect what failed, what changed and what passed.
- 01
Request
Return 20% for $6M raised at $30M post-money.
- 02
Execution
CaseFile captures real Node assertion failure and corrected execution.
- 03
Decision
Eval checks the correction. Accept this fixture; no production release.
Inspect the correction
− raise / (postMoney + raise) * 100 + raise / postMoney * 100Open failure, commands & evaluation →
Adopt the ecosystem before the language.
Keep application code in its existing language. Introduce accountability at the workflow boundary.
- 01
Today: wrap the task
Use CaseFile for command evidence and Eval for explicit repository checks.
- 02
Then: control execution
Adopt Dispatch or Workcell after validating the target environment and runtime pins.
- 03
Over time: go deeper
Choose Kujo for new agent tooling where native language integration adds value.
Start with teams already delegating engineering work.
Proposed beachhead: agent-platform builders shipping repeatable engineering workflows.
- 01
Buyer / user
Platform engineering lead / engineers operating coding-agent harnesses.
- 02
Immediate pain
Failed runs are expensive to reconstruct. Review spans commits, logs and agent transcripts.
- 03
Purchase hypothesis
Pay for retained evidence, controlled execution and team review if they reduce investigation time.
Open tooling can lead to a commercial operating layer.
Earn developer trust through useful local tools. Charge for shared operational responsibilities.
- 01
Available foundation
Open repositories and local developer tooling.
- 02
Proposed team service
Hosted evidence, retention, shared review and execution capacity.
- 03
Proposed enterprise tier
Organization policy, deployment controls, support and governance.
A focused entry market can support a meaningful business.
Illustrative bottom-up model. Organization counts and annual contracts are assumptions to validate.
- initial target organizations × $24K ACV
- 2,000
- illustrative annual serviceable revenue pool
- $48M
- 100 customers × $24K; milestone scenario, not forecast
- $2.4M
Accountability requires more than generating code.
Existing platforms are strong. Kujo must win on useful evidence and the cost of operating reliable workflows.
- 01
Coding agents / GitHub
Generation, repository context and review are already integrated into major developer surfaces.
- 02
Orchestration / LangGraph
Persistence, recovery and human review are established capabilities.
- 03
Kujo opportunity
Connect explicit contracts, bounded work and inspectable artifacts. Prove adoption before claiming a moat.
Meet developers inside their existing workflows.
The planned December 2026 open-source campaign should produce repeat usage and customer learning.
- 01
Discovery
GitHub releases, technical demos, docs and reproducible fixtures → qualified visits.
- 02
Activation
Agent harness recipes and MCP tooling → first successful evidence run.
- 03
Retention → revenue
Repeat weekly runs → design-partner discussions → paid pilots and renewal evidence.
Execution is visible. Demand still needs proof.
Engineering output, external adoption and commercial traction are different measures.
- 01
Engineering output
Public repositories, documented releases and the locally reproduced evidence fixture.
- 02
External adoption
Verified outside usage, retention and independent integrations: data not supplied.
- 03
Commercial traction
Design partners, paying customers and revenue: data not supplied. Complete before investor distribution.
Built through sustained open-source engineering.
Robert DeVore connects open-web product work with a growing language and developer-tooling ecosystem.
- 01
Background
Public biography describes years of WordPress, WooCommerce, plugins and publishing systems.
- 02
Current work
Language/runtime infrastructure, orchestration, evaluation, telemetry and evidence workflows.
- 03
Team enabled by capital
Proposed hires in runtime, execution reliability and developer experience; later customer engineering.
Capital turns a technical foundation into an adopted platform.
Working seed scenario. Financing terms and hiring pace remain assumptions.
- target raise
- $6M
- post-money valuation or SAFE cap assumption
- $30M
- planned runway, including reserve
- 24 mo
0–6 mo
Flagship evidence workflow6–12 mo
External validation + pilots12–24 mo
Paid retention + repeatability
Flagship evidence workflow6–12 mo
External validation + pilots12–24 mo
Paid retention + repeatability
Agents will build more software.
The organizations deploying it will still be accountable for what it does.
KUJO
Trust at the speed of AI.Investor conversations ↗
Trust at the speed of AI.Investor conversations ↗