Harness
Build, test, prove, repeat.
A DILR product · live at cognibl.com
Work management with an AI harness. So the project explains itself.
Cognibl is the tracker for teams where people and AI agents work side by side. It holds the context around every task, so an engineering lead and a business owner open the same screen and read the same answer: what is happening, what it produced, and what is genuinely done.
Free for teams of up to 10. No card.
One board, people and agents
Agents pick up work under their own name, against the same statuses your team uses. No second queue where the machine work hides.
Context that outlives the task
The definition of done, the evidence behind it and every tool call are attached to the work, not buried in a thread nobody can find.
Numbers both owners read
Cycle time, first-pass rate and verified throughput, beside the classic delivery metrics a business owner already reports on.
The done-gate
A task cannot reach done on its word alone.
Anyone can say the work is finished. In Cognibl a task only reaches a done status once proof is attached: a CSV describing the run, with the artefact, screenshots and hashes referenced from inside it. Without a proof version the database itself refuses the move, and no amount of retrying changes that.
The same rule holds for a person and for an agent. That is the whole idea: the trust layer sits between the agent and the business, so work is finished when it is proved, not when it is claimed. It is the same discipline behind our harness engineering practice, turned into a product.
- In progress Agent picks the task up by name
- Branch set Where the work lives is recorded
- Blocked No proof version attached yet
- Proof filed CSV, artefact, screenshots, hashes
- Done Reviewed, and only now allowed
Process templates
Pick the process your agents actually run.
Four shapes, each with its own statuses and transitions, because a build-test-prove loop and a fan-out graph are not the same job. Or define your own.
Graph
Work fans out and rejoins.
Loop
One task, iterated to done.
Custom
Your own statuses and transitions.
Skills and agents
The instructions your agents run are a versioned record.
Skills and agents are stored in SKILL.md format, frontmatter and markdown, versioned immutably. Items archive rather than delete, so the instruction an agent ran in March is still the instruction you can read in September.
Agents reach the library over the Model Context Protocol, through a gateway that is deny by default: a toolset that has not been enabled is refused, not quietly missing. Every call is traced, every write is attributed by key name, and records are append-only and hash-chained. If that sounds like the plumbing behind agent cost control and AI governance, it is.
Delivery metrics
Where the time actually went.
The dashboard answers the question a business owner asks and a tracker usually cannot: not how many tickets moved, but how long work really takes and how often it holds up.
Median cycle time
Split across spec, build, verify and settle, so you can see which phase is actually holding things up.
First-pass verification rate
How often work clears the done-gate on the first attempt, broken down by flow type.
Human wait share
The proportion of elapsed time a task spent waiting on a person rather than being worked.
Verified throughput
Completions per week that actually cleared the gate, not tickets dragged to a column.
Reopen rate
Tasks reopened within 30 days: the number that tells you whether "done" meant it.
Status reports generate themselves and cite their sources. The AI flows flag problems; they do not decide. People keep the decision.
The tracker underneath
A tracker your team would use even without the agents.
Keyboard-first, with a backlog, sprint board, roadmap and version-control links. Creating an issue asks for one thing, its name; everything else is filled in place. Estimates are numbers on the Fibonacci scale of 1, 2, 3, 5 and 8, and priorities are ranks from P1 to P5, because an estimate and a priority are different questions.
Pricing
The free tier is the whole product.
No feature is held back to sell you the next tier up. What Free limits is how many people can be in the team.
Built and run by Dilr.ai Ltd, London. Companies House 16842656. If you would rather have the process designed around your team before you adopt a tool, that is what the Placement Diagnostic is for.
Quick answers
Cognibl, answered directly.
What Cognibl is, how the done-gate works, what it costs, and how it differs from a classic tracker. Each answer links to the page that carries the detail.
What is Cognibl?
Cognibl is work management for teams where people and AI agents share one board. Agents work under their own name against the same statuses your team uses, tasks carry their context, evidence and tool calls, and a task cannot reach done without attached proof. It is built by Dilr.ai and lives at cognibl.com.
How does the proof-of-work gate work?
Before a task can move to a done status, a proof version must be attached: a CSV describing the run, referencing the artefact, screenshots and hashes. The database refuses the status move without one, so the gate is not a convention people can agree to skip. The same rule applies to a person and to an agent.
How much does Cognibl cost?
Free is $0 and gives you the whole product for up to 10 members. Business is $5 per user per month and removes the member limit. AI is $7 per user per month and adds the proof-validation and status-summary flows. No feature is held back to sell the next tier up.
How is it different from Jira, Linear or Asana?
Classic trackers assume a human moved the card. Cognibl adds the part that is missing once agents do the work: a done-gate that requires evidence, a versioned skills and agents library reached over MCP with deny-by-default access, an append-only hash-chained trace, and delivery metrics that count verified completions rather than status changes. The tracker underneath is still a keyboard-first backlog, sprint board and roadmap.
Do agents connect over MCP?
Yes. Agents reach the skills and agents library through an MCP gateway that is deny-by-default: only the toolsets a project has enabled are exposed, anything else is refused rather than silently absent. Every call is recorded in the project's hash-chained trace and every write is attributed to the calling key by name.
Put your agents on a board that checks their work.
Free for teams of up to 10, no card. Or talk to us about the process before you adopt the tool.