[ 01 ]
Evidence you can inspect.
Explore your material, ask questions, and follow citations back to the source. Keep observations distinct from assumptions.
LOCAL RESEARCH & DELIVERY WORKSPACE
Your next move.
Backed by
evidence.
A local workspace for turning recurring feedback and research into reviewed agent work. Keep the sources, method, decision, and verification connected—even when the evidence changes.
LOCAL RECORDS. EXPLICIT REVIEW. YOUR AGENT.
[ THE CONTEXT STAYS CONNECTED ]
Customer feedback 01
“I signed up. What do I do next?”
Product research
A pattern, not just
a single comment.
Make sense.
Make a decision.
Make a decision.
Agent-ready work For review
Make the first step obvious
↗ Evidence✓ Acceptance criteria
01 / SIGNAL → ACTIONILLUSTRATIVE WORKFLOW
Better decisions start
with the whole picture.
◌ Customer feedback⌕ Market research▧ Competitor insights≡ Your backlog
01 — THE WORKFLOW[ INTERACTIVE ]
Signals come in.
Clarity comes out.
Follow a sample problem through evidence review, a bounded decision, an agent handoff, and verification. Inspect the reasoning at each step.
Interactive product story · illustrative data
One decision, from selected evidence to a verifiable change.
HUMAN JUDGMENT AT THE CENTRE.THE CASE THAT MATTERS — ROUND TWO
The source changed.
What happens to the task?
Recurring research brings corrections, contradictions, and new constraints. Tonelingo tracks their effect on existing work instead of treating each import as a fresh assignment.
SYNTHETIC EXAMPLE · SOURCE 03
“Every import fails, even with valid rows.”
The initial report leads to a proposed import-validation fix. A human approves task version 1.
DECISION D-014Initial review
EvidenceSource revision 1
TaskVersion 1 · investigate import validation
ApprovalBound to version 1
Next handoffUses the approved scope
The record keeps the selected evidence, rationale, scope, and approval together.
This illustrates the implemented revision and invalidation model. It does not execute an agent or certify a real task.
Read the engineering case study ↗02 — THE CONNECTED WORKSPACE
The context should
travel with the work.
The same decision record links source revisions, the research review, a versioned task, its approval, and the result. Each transition has explicit checks.
[ 02 ]
A task with explicit scope.
Translate product intent into structured tasks. Carry the reasoning, requirements, and boundaries into the next handoff.
[ 03 ]
Keep the why alive.
Source corrections invalidate dependent approval. Revisit affected work while retaining its history and the context that still applies.
03 — YOUR AGENT. BETTER CONTEXT.
You choose
the builder.
Bring the why.
Tonelingo binds the approved task to its version, contents, target repository, and base commit. A stale handoff must be reviewed again before execution.
Native Codex and Claude Code adapters are implemented for manually initiated work. Real authenticated provider runs remain a validation requirement; an export can also be reviewed in your existing tools.
See an example handoffTHE INTENT, IN ONE PLACE
FROM TONELINGO
One clear brief.
The problem. The evidence.
The scope. The success criteria.
EXPORT & HANDOFF
Your existing coding agent
Continue in the tools you use.
You initiate execution. The agent uses its own environment and account. Completion requires a separate verification record for the exact result commit.
04 — RESEARCH YOU CAN INSPECTMETHOD → EVIDENCE → JUDGMENT
The method belongs
in the record.
Specialist playbooks define the questions, labels, counterexamples, and review requirements. A saved review retains the exact method and codebook used to reach the decision.
CUSTOMER FRICTIONHuman-reviewed method
Which friction warrants a change?
Required context
Selected feedback, the affected journey, and the current product behavior.
Research judgment
Code the evidence, inspect contrary reports, and explain the proposed disposition.
Possible conclusions
Make a change · investigate · use an existing solution · do nothing · unknown.
Retained with the decision
Method and codebook snapshots, reviewed records, form answers, rationale, and reviewer.
USE YOUR CHOSEN RESEARCH TOOL
A reviewable exchange.
Export selected evidence and the chosen playbook as a research packet bound to an exact SHA-256 hash. Use an external analysis tool, then import its structured suggestions into the existing review.
01
Export selected evidence
Explicit contents and method
02
Check returned suggestions
Known sources, codes, and matching quotations
03
Review before accepting
Editable proposals; no automatic approval
Current exchange uses manual files. Local heuristics are limited; a matching quote establishes text correspondence, not a correct interpretation.
Research exchange contract ↗THE RESEARCH QUESTION
What changes when
the evidence changes?
The work should know.
SOURCE REVISION → REVIEW → VERIFIED CHANGE
05 — ENGINEERING & EVALUATION
Concrete boundaries.
Testable behavior.
The local application uses one Next interface, API, and persistent FileStore. Implementation, executed checks, provider behavior, and observed user value are recorded separately.
Your local records.
Project state persists in .tonelingo2-data/. Portable decision reports retain the reasoning and revision history. Full backups and project exports serve different purposes.
The application does not encrypt the local FileStore. Local storage is not a guarantee of fully offline processing.
Explicit data exchange.
Basic local import and exploration need no provider keys or SPEEQO account. External research receives the contents you choose to export.
Using a cloud model or native agent may transmit context to its provider. Its privacy and account terms still apply.
Review before action.
Approval binds to a concrete task and repository state. The user initiates a native agent attempt; independent checks bind to the exact result commit.
Unattended workers, live connectors, and hosted multi-user operation are outside this preview.
THE ADVANTAGE WE ARE TESTINGLess human reconciliation
Less human reconciliation
when evidence changes.
The comparison must include a competent existing workflow: ordinary files, research tools, a specification toolkit, and a coding agent. Extra review can cost more than it saves on a one-off task.
01
Total human effortCount setup, preparation, review, rework, and support—including failed or abandoned attempts.
02
Integrity after a correctionCheck that affected work requires renewed approval and unrelated work remains stable.
03
Accepted work & repeat useCompare independently reviewed quality, retain failures, and observe whether people return for another real task.
Research hypothesis
No measured productivity advantage or independent adoption result is claimed. The pilot protocol defines how to test it.
Inspect the evaluation protocol ↗06 — WHERE IT FITS
Build on capable tools.
Study the handoffs.
Research, templates, agent choice, and approval are established capabilities. Tonelingo’s focus is the continuity of a decision through recurring evidence and code changes.
| Tool / project | Documented strength | Tonelingo’s focus alongside it |
|---|---|---|
| ChatGPT Deep Research ↗ | Multi-source investigation and structured reports with citations. | Keep reviewed findings linked to a persistent decision and the work affected by later source revisions. |
| GitHub Spec Kit ↗ | Specifications, plans, tasks, implementation, agent integrations, templates, and extensions. | Track how source corrections affect existing research, task revisions, and the validity of an approval. |
| Claude Code ↗ | Codebase analysis, implementation, testing, and Git workflows. | Bind the handoff to approved intent; retain independent commit verification and subsequent outcome observations. |
A comparison of documented capabilities, not an exhaustive feature audit or proof that another tool cannot support the same process. Whether the connected workflow saves effort is an empirical question.
07 — RUN & EXPLORELOCAL SINGLE-USER PREVIEW
Start with one
real research question.
Use feedback you are authorized to analyze. Start with a small recurring problem and an existing codebase, so you can judge the value of carrying the context forward.
LOCAL SETUP · ACCESS REQUIRED
git clone https://github.com/Speeqo/tonelingo2.git cd tonelingo2 corepack enable corepack pnpm install --frozen-lockfile corepack pnpm build:all corepack pnpm dev:local
Then open
http://127.0.0.1:3000API: 127.0.0.1:8787BEFORE YOU START
A local development setup.
- Repository
- GitHub access is currently required.
- Runtime
- Node 22.14+ within 22.x, or Node 24. Initial targets: Linux and macOS.
- Package manager
- pnpm 10.0.0 via Corepack.
- Services
- No Postgres, Redis, provider key, or SPEEQO account required for the basic local workflow.
If Corepack is unavailable, install pnpm 10.0.0 and use pnpm directly. For offline setup diagnostics: node scripts/oss-doctor.mjs.
- 01
Import & inspect
Import CSV or JSON in a project. Check source identities, mappings, privacy choices, and the selected records.
- 02
Research & decide
Ask a grounded question in Workspace. Follow the citations, prepare a change, and complete the research review.
- 03
Review & hand off
In Develop, set the repository, base commit, allowed files, and verification commands. Approve the exact task revision, then initiate the chosen agent handoff.
- 04
Verify & revisit
Review the actual diff and independent checks. Attach observations when available, then introduce a corrected source and review the affected task.
Start with an empty local project. Synthetic demo records are opt-in.Supported release scope ↗
08 — PROJECT & KNOWLEDGE BASE
Inspect the design.
Follow the evidence.
Tonelingo is an engineering and research project led by Dr. Petra Kolesarova, co-founder and CEO of SPEEQO. The repository documents the architecture, tradeoffs, evaluations, and remaining work.
Speeqo / tonelingo2Community preview preparation. The repository is currently private; links require access. A public open-source release and contribution terms are pending.
Source reviewed 14 Sep 2026
main @ b74b7b9c ↗DESIGNEngineering case study
State transitions, competing approaches, and the falsifiable product thesis.
↗INSPECTABILITYPortable decision reportSources, rationale, approval, and verification in one reviewable record.
↗STATUSImplementation & remaining evidenceImplemented local behavior, unobserved use, and the release boundary.
↗EXTENSION MODELSpecialist playbooksExtend research questions, labels, counterexamples, and review requirements. External contributions await finalized terms.
↗SCOPE & PRACTICAL QUESTIONS
Know the boundaries.
What is Tonelingo?
Tonelingo is a local research and delivery workspace for recurring product evidence. It connects source records, an explicit research review, a versioned decision and task, a reviewed agent handoff, and verification. The target users are software teams and studios already working with customer feedback and coding agents.
Does Tonelingo replace my coding agent?
Your coding agent performs the implementation. Tonelingo prepares and tracks the reviewed task and its evidence. Native handoffs are manually initiated, and the agent’s report is followed by independent checks tied to the actual result commit.
What is available today?
The current local scope includes CSV/JSON import, cited exploration, selectable research reviews, source and task revisions, approval invalidation, native handoff adapters, independent verification records, and portable reports. Actual authenticated provider rehearsals, independent user value, and public release readiness remain separate requirements. The examples on this website use synthetic data.
Can I connect all my tools?
The supported preview uses local imports, exports, and manually initiated native handoffs. Live third-party connectors, automatic Jira/GitHub delivery, hosted billing, and unattended multi-agent execution are outside this release. The setup and release guides describe the current boundary.
RESEARCH → REVIEWED CHANGE
Bring a specific question.
Leave with inspectable work.
SPEEQO offers scoped research and implementation assistance:
review authorized evidence, prepare a bounded change, and leave the records with your team.
SCOPE, FEES, AND RESPONSIBILITIES AGREED PER ENGAGEMENT.