Observability · Colony Ledger

The Ledger.

Every prompt, every tool call, every dollar — attributed to the issue, the agent, the repo, and the tenant. Queryable Postgres. CSV export. The answer to “what did AI actually cost us this month” is a SQL query, not a guess.

What it is

Cost attribution by construction.

The Ledger is not a report we generate at month-end. It is the source of truth for every agent invocation, written transactionally as the work happens. The dashboard, the export, and the rollups are all projections of the same underlying log.

$

Per-issue.

Every issue ID carries through the pipeline. Every prompt, completion, and tool call posts back against it. The CFO question “what did this PR cost” has a one-row answer.

Per-agent.

Builder, Inspector, Architect, Surveyor, Marshal, Mayor, Sentinel — spend rolled up per role, with shares. You see which agents are doing the work and which are spinning.

Per-tenant.

Multi-team and multi-repo deployments roll up cleanly. Per-tenant ledgers are isolated; cost attribution survives reorgs and BU splits because it is keyed on the tenant, not on the dashboard view.

With and without

What “the AI bill” looks like.

Without Colony

A line item on the model-provider invoice.

  • One number, by provider, by month.
  • Not attributable to a feature, repo, team, or issue.
  • Spikes are unexplained; spend caps are out-of-band.
  • Reconciling to engineering output requires a spreadsheet exercise.
  • CFO asks “what did the AI agents cost us” and the answer is “roughly this many dollars.”
With Colony Ledger

$0.91 median per merged PR. Attributed.

  • Per-issue, per-agent, per-tenant rows, written transactionally.
  • Rollups by feature, repo, team, sprint, or any tenant-defined dimension.
  • Per-issue cost caps enforced by the pipeline; spikes are predictable.
  • Finance reconciles to the provider invoice within rounding.
  • CFO asks the question and the answer is a SQL query.
Sample export

The actual CSV a finance team gets.

Below is the first few rows of the sample download. Same columns, same types, same shape your tenant’s Ledger produces. The unit-price column and the markup column are real columns; there is no hidden multiplier.

issue_id,repo,merged_at,primary_agent,state_path,tokens_total,cost_usd,duration_min,human_review
2841,RunColony/colony,2026-05-13T11:42:00Z,Builder,"intake→survey→draft→build→inspect→marshal→merged",412184,1.84,12,false
2840,RunColony/colony-cloud,2026-05-13T10:18:00Z,Inspector,"intake→survey→draft→build→inspect→marshal→merged",96042,0.42,7,true
2839,RunColony/colony,2026-05-13T09:54:00Z,Surveyor,"intake→survey→draft→build→inspect→marshal→merged",128331,0.51,9,false
2838,RunColony/colony-cloud,2026-05-13T08:41:00Z,Architect,"intake→survey→plan→draft→build→inspect→marshal→merged",204120,0.88,15,false
2837,RunColony/colony-cloud,2026-05-13T07:22:00Z,Sentinel,"intake→survey→draft→build→inspect→marshal→merged",66218,0.28,6,false
Download the full sample CSV →
For procurement

The questions every finance team asks. Answered.

Bring more on the pilot call — we answer specifically and we don’t hand-wave.

Q How do you attribute LLM token spend back to a specific issue?

Every agent invocation carries the issue ID through the pipeline; each prompt and each completion is logged against that ID with the provider, the model, the input tokens, the output tokens, and the unit price at the time of call. The same row records the agent that issued it. Rollups by issue, by agent, by repo, by tenant, and by month are precomputed views over the underlying log.

Q Can finance get an export they can reconcile against the model-provider invoice?

Yes. The Ledger export is a CSV with one row per agent invocation, including tokens and unit price; sums by provider and month line up with the provider invoice within rounding. The sample export linked above is the same shape your finance team would receive.

Q Do you mark up tokens? Where is the markup visible?

Token cost is pass-through at the provider unit price plus a small markup, set on the contract and visible as a separate column in the Ledger. Worker capacity, platform tier, and managed services are priced separately. There is no place where Colony cost is opaque.

Q How long is data retained, and can we delete it?

The Ledger is retained for the period required by jurisdiction for finance and tax purposes; prompts and completions can be deleted on request within thirty days with a signed deletion certificate, separate from the cost rollups required for audit.

Q Can we run our own dashboards against the data?

Yes. The Ledger is queryable Postgres views with a stable schema. Wire it into your BI tool, alert on spend per agent, build a dashboard your CFO actually reads. The point is that the data is yours.

See the Ledger against your repos.

The pilot is six weeks. We pick a slice of your work; the Ledger picks up from there.