Skip to content

Assign tasks to cloud AI agents

Cognibl runs coding agents on Claude Code and Codex, right on your team's board. Each one opens a pull request with proof of work, and only a person can merge.

Assign your first taskHow cloud agents work

Free for one person and $5 per seat for teams. Agents draw prepaid credit at a published rate for each model.

Cognibl product walkthrough: one week of agent work. A five minute walkthrough of Cognibl at a demo company, Quillfern Coffee. People and cloud agents share one board, and every task carries a definition of done. An agent reconciles supplier invoices against a bank export and files the result as proof of work. Another agent fixes a checkout bug and opens a pull request, a review agent checks it against the definition of done, and a person merges. Every agent call is recorded in a hash-chained trace, recurring tasks run on a schedule, and no task is completed without proof.

Cloud agents

Assign work to people and agents from the same list

Pick an agent like you pick a person. It returns a pull request and proof of work.

Explore cloud agentsStart free

Assignee

  • Maya Chenperson
  • Nimbusagent · starts a run
  • Atlasagent · starts a run
  • Orionagent · out of credit

Connections

Agents can read your tools but never hold your credentials

Connect 94 tools once. Each project picks what agents can read, and every call is logged.

See every connectionStart free

Tools for agents

3 granted
  • GitHub
  • Google Workspace
  • Linear
  • Atlassian
  • Notion
  • Microsoft Learn

Read-only by default. An admin can let agents write to Google Workspace.

Recurring tasks

Schedule repeat work and get proof every time it runs

Write it once and set a schedule. Every run is checked like any other task.

How recurring tasks workStart free

Summarise yesterday's failed payouts

Daily at 08:30 Europe/London · Atlas
  1. Mon
  2. Tue
  3. Wed
  4. Thu
  5. Fri
  6. Sat
  7. Sun

Version control

Pull requests anyone on the team can follow

A branch and pull request per task, with one clear status. Only people merge.

Explore version controlStart free

Pull requests

  • Ready to mergePAY-21
  • 2 checks failedPAY-19
  • Changes requestedPAY-22
Only a person can merge.Merge

How work moves

Choose how work moves through each project

Harness, Graph and Loop come ready. Pick one per project.

  • Harness

    Build, test and prove until done.

  • Graph

    Work fans out and rejoins.

  • Loop

    One task repeated until it is done.

  • Custom

    Your own statuses.

First-pass verification rate

Illustrative

The share of agent tasks that pass on the first try. A clear definition of done passes the first time. A vague one shows up here before it reaches production.

First-pass verification rate by flow, as a percentage.
FlowFirst-pass rate
Harness82%
Graph64%
Loop71%
Custom48%

Proof of work

Nothing is done until the proof is attached

Attach a CSV, its documents and a coverage report. Without them, done is refused.

See how proof is checked
TEST-2Update schema of user authentication tables
In reviewDone

Refused. This status requires a proof of work.

The harness template expects a CSV and a coverage report. Neither is attached to this task.

Attach proof of work

Skills and agents

Agent instructions are versioned and never overwritten

Skills and agents live in each project as SKILL.md files.

Skills

Instructions an agent reads before a type of work.

Branch disciplinev4a91f3c2

Agents

A manifest, an optional model and tool list, and a system prompt.

Schema reviewerv2a91f3c2

  • Every edit is a new version

    Edits make a new version. Nothing is deleted.

  • Start from our catalog

    Install from ours. Updates wait for your yes.

  • Agents fetch them over MCP

    Off until you allow it. Every call traced.

Issues

A complete work tracker before you add any agents

Backlog, sprint board, roadmap and version control in one place. It is fast, works from the keyboard and adds no extra process.

Sprint 2

Running26 Aug to 8 Sep
  • INVO-13Chase the missing May credit notesP3
  • INVO-12Reconcile the April supplier ledgerP3
  • INVO-7Split the April credit notes by supplierP3
  • INVO-6Wire the reminder emails into the pipelineP3

7 more in the backlog

Creating an issue asks for one thing

Update banner image on landing pageP3Open
Issue

Type a name and press enter. Add priority, status, estimate and branch later, one at a time.

Priorities and estimates are numbers

P1P2P3P4P512358

Priority runs from P1 to P5, so nobody argues whether urgent beats high. Effort uses 1, 2, 3, 5 and 8, and the burndown counts the same number you see.

Metrics

See where the time goes on every project

Cycle time split into spec, build, verify and settle. Writing the spec often takes longest.

Median cycle time

Illustrative figures
Median cycle time by project, in hours, split into Spec, Build, Verify and Settle.
ProjectSpecBuildVerifySettleTotal
Invoicing31h18h9h3h61h
Onboarding14h22h6h2h44h
Data import44h12h17h4h77h
Reporting9h26h5h2h42h

Long spec time plus a low first-pass rate means an unclear spec.

  • First-pass verification rate

    Agent tasks passing on the first try.

  • Human wait share

    Time spent waiting on a person.

  • Verified throughput

    Verified completions per week.

  • Reopen rate

    Reopened within 30 days.

Raw task counts and signups are left out on purpose. They measure effort, not delivery.

For the people running the project

AI writes the status report and links to its sources

Two AI flows summarise and flag problems. They never decide.

What it caught on this roadmap

13 problems
  • Pilot cutover has moved 39 days later than the plan it was committed against.
  • Five milestones have no work on them, so their progress cannot be measured.

A milestone with no work under it is not on track. Its progress is unknown.

Proof validation

On every status change, a flow reads the attached proof of work against the definition of done and writes a summary of what matched and what did not. The result is saved on the task as a record.

Project status

A second flow reads the week's tasks, proofs and traces and writes the summary at the top of the project homepage. Nobody has to put it together by hand before a review.

Compared

How Cognibl compares with a classic tracker

A full project tool first. This is what it adds.

Cognibl compared with a classic issue tracker, across six capabilities.
CapabilityA classic trackerCognibl
Moving a task to donePick done from a dropdown. No evidence needed.Refused without a proof of work attached.
What an agent may touchWhatever the API token it was given allows.Only what the policy map names. Everything else is blocked and logged.
The record of what happenedWhatever the integration chose to report.Written by the gateway, append-only and hash-chained.
ProcessColumns you name yourself.Harness, Graph and Loop come ready, and only allowed moves are possible.
MetricsVelocity and burndown, counted from completions.Cycle time by stage, first-pass rate, human wait share, counted from verifications.
Instructions an agent ranA prompt in someone's editor.A versioned library item that never changes and can be looked up months later.

Pricing

Start free and upgrade when you add people

Creating an organization is free. Business features stay visible and locked until you upgrade.

Personal
$0for one personYour own workspace, your projects and your agents.
Business
$5per seat, per monthAdds roles, single sign-on, an audit log, shared context and a brand kit your agents can read.
Start freeCompare plans and read common questions