Kilo Code

Kilo Code is the most complete free bundle here: an editor agent, a CLI and a JetBrains plugin under one roof, plus cloud agents when you want to hand a task off to a container and come back later. The trait that decides it is coverage: one config, one login, three surfaces, and the same modes in each.

The cost of the bundle is that each surface inherits its upstream's quirks, so the CLI and the extension do not always behave alike. Pick it for a mixed team that refuses to standardize on one editor. Pick Cline if you want the upstream agent and nothing else.

The lineage is the risk. The extension and the CLI are each forks of a different upstream, so two upstreams means two streams of changes to merge or diverge from, and each divergence becomes a bug nobody upstream will see. That is manageable while the merges keep coming and expensive the day they stop.

The consequence for a user is that a fix upstream reaches Kilo only when someone merges it, so watch the merge cadence, not the release notes. What it does right: Cloud Agents run in isolated Linux containers, which neither upstream ships, and which gives a runaway task somewhere safe to run.

Kilo's design is inherited. 1. The editor agent's context gathering and edit application are Roo Code's. 2. The CLI's provider abstraction is OpenCode's. 3. The original element is Cloud Agents in isolated Linux containers, which adds a verification boundary neither upstream has: a task can run its own tests without touching the developer's machine.

No benchmark is published for any surface. The consequence is that the quality of the loops is whatever the upstreams' quality was at the last merge, plus the sandbox. The observation: forking two good designs and adding a sandbox is a reasonable way to build, if the merges continue.

Kilo already has an owner: Anaconda, July 2026, license intact. That settles the 18-month question and opens another, because Anaconda's distribution is Python and data teams, and a coding agent is adjacent to that, not central. Adjacent products inside an acquirer get the budget the core does not need.

Moat: none of its own; the moat is now the acquirer's sales channel into enterprises that already run Anaconda. The pivot to watch is the gateway, where the margin lives. Position: safe for 18 months, expect the free tier to be revisited first and the enterprise tier to be pushed hardest.

The demo is a familiar editor agent plus a CLI. Procurement: Teams is $15 per user, so sixty seats is $900 a month before gateway usage, the cheapest seat here that covers both VS Code and JetBrains, so the two editor camps get one invoice. The acquirer is a vendor procurement may already know from the data team.

CI fit is real, the CLI runs headless, and onboarding is an extension install with a shared config. Approved with conditions: SSO and retention confirmed in writing, and gateway spend capped per team, because a pay-as-you-go gateway is a meter with sixty hands on it.

I can read the editor half and the terminal half, because they are forks of projects I already read, and the diff against upstream is the changelog. Local models through Ollama or LM Studio are documented, and the MCP config is a JSON block where I set command, environment and a timeout per server, which most tools forget.

The phrase I will re-read every quarter is remains MIT, because acquisitions change licenses after the press release. A fork survives regardless, since two upstreams already exist to fall back on. Respect, not grudging, and a bookmark on the license file.