Releases
RSSv0.83.0
- **The settings search finds the Claude bridge switch.** Searching "Claude", "bridge" or "operate" (in every language the app offers) now lands on the "Let Claude operate this app" switch and on full control, opening the Advanced block and moving focus to the switch. The other A
v0.80.0
- **Skill card: do not widen types to silence the checker.** A new pattern card on narrowing a value or handling the missing branch instead of widening an annotation to `Any`, `object` or an unchecked union, translated into every site language. Contributed by @CG-5228. (#1150)
Chimera 0.79.0
Chimera 0.79.0 ## [0.79.0] - 2026-10-27 ### Added - **Connect VoiceStudio in one click.** Settings › Connectors lists "VoiceStudio (local)", a local voice studio (cloning, voice design, speech, transcription). The card connects when VoiceStudio is running and says how to install
Chimera 0.78.2
Chimera 0.78.2 ## [0.78.2] - 2026-10-27 ### Added - **SVG images are accepted, after validation.** Vector models such as Recraft V4.1 Vector return SVG; Chimera rebuilds it from an allowlist (no scripts, event handlers, external links or entity declarations), saves it as `.svg`,
Chimera 0.78.1
Chimera 0.78.1 ## [0.78.1] - 2026-10-26 ### Fixed - **Image-only models work through OpenRouter.** FLUX, Recraft, GPT Image and Krea go to OpenRouter's images endpoint with the size parameters each model declares; before, every one of them failed instantly with a bare "image gene
Chimera 0.78.0
Chimera 0.78.0 ## [0.78.0] - 2026-10-26 ### Added - **Chimera Design draws as the agent works.** The canvas applies the agent's changes batch by batch instead of reloading the whole design, falls back to a full reload when a batch is missing or fails, and shows "Creating…" while
Chimera 0.77.0
Chimera 0.77.0 ## [0.77.0] - 2026-10-25 ### Fixed - **The desktop app's background service no longer gets stuck half-stopped.** A messaging adapter that had never connected (for example a Discord bot with a wrong token) made the service's shutdown fail partway. The service then s
Chimera 0.74.2
Chimera 0.74.2 ## [0.74.2] - 2026-10-23 ### Fixed - **With the verify-and-repair loop on, a turn no longer waits an hour after its answer in a large repository.** The "changed after it passed" check read every file of the project one git call at a time; it now compares only the p
Chimera 0.74.0
Chimera 0.74.0 ## [0.74.0] - 2026-10-22 ### Changed - **Cutting a release now bumps the Chrome extension's version too.** The manifest, its README and the extension docs move to the new version with the package, so the store always receives a higher version. (#1066) ### Fixed - *
Chimera 0.73.0
Chimera 0.73.0 ## [0.73.0] - 2026-10-22 ### Changed - **The Chrome extension's popup now looks like part of Chimera, in light and dark.** It used to show bare browser buttons; the connection status, the button that allows a site, Stop and the pairing code are now laid out and the
v0.71.0
Chimera 0.71.0 ## [0.71.0] - 2026-10-21 ### Added - **Foundation for browsing in your own Chrome.** A read-only Chrome extension and the engine behind the existing browser tool, so pages read there get the same fence, taint and site list as today. It is not connected to anything
v0.68.1
- **The desktop app starts again.** In 0.68.0 its backend died at start on every machine with `No module named 'chimera.cli.commands.apilogs'`: the CLI imports its command areas by name, PyInstaller follows only import statements, and the freeze left that area out. The freeze now
v0.67.3
- **Discord webhook delivery supplies an explicit User-Agent.** Discord's Cloudflare returned HTTP 403 with error 1010 for urllib's default User-Agent, preventing webhook messages from being delivered.
v0.67.2
- **Bot MCP access is an explicit server allowlist.** `CHIMERA_MCP_AUTOLOAD` no longer exposes its connected servers to messaging bots by itself; `CHIMERA_BOT_MCP_SERVERS` opts named servers in, and every rea tool call that runs there is reported to the owner like on any other su
v0.67.1
- **MCP tool responses can be capped per server.** Set `max_output_chars` or use `--max-output-chars` with `chimera mcp add` to keep a large response from overwhelming the prompt; truncated output reports its original size.
v0.67.0
- **Document the rea MCP server.** Add its catalogue setup, unattended tool restrictions, prerequisites and timeout configuration to `docs/rea.md`.
v0.64.5
- **An approval's resolution is chained into the audit log, and the log is fenced from the agent's own shell** (#783). `approvals/history.jsonl` sits where the agent's shell can reach, so a run could rewrite the record of the question that governed it. Every durable approval now
v0.64.4
- **A value saved to `.env` reads back as exactly that value, through every writer** (#775). This holds whether or not the desktop bridge is on: the owner's own save, key pools, connectors, vault moves and `chimera init` all went through the same gaps. - **Line breaks** (shipped
v0.64.3
- **Settings that widen what the agent reaches, or loosen a guard or a privacy fence, are the owner's alone** (#770, #772). Through the desktop bridge's `settings.edit`, a client at full control could set its own posture (`CHIMERA_REACH`, `CHIMERA_APPROVAL`, `CHIMERA_HOST_EXEC`),
v0.64.2
- **Every chat bot answers only the ids its owner listed** (#746). The Discord, Telegram, Slack and Signal adapters accepted an allowlist and nothing filled it, so a bot answered anyone who could write in a channel it read, with the owner's tools and spend. `CHIMERA_DISCORD_ALLOW
v0.64.1
- **A slow Intel dmg no longer holds back the update of every platform** (#732, in this tag; listed under Unreleased in the CHANGELOG). - **What was wrong:** on the `macos-15-intel` runner, Tauri's `bundle_dmg.sh` went from 1.5 minutes (v0.61.2) to failing after 10 (v0.64.0, twic
v0.64.0
v0.63.1
- **The desktop agent can start long commands in the background, read them and stop them, inside `run_shell`'s fences.** A bench stage driven through the MCP bridge was cut at 120 s. The existing `run_shell(background=true)`, `job_status` and `job_cancel` now close the gaps: - **
v0.63.0
- **Answers grounded in your documents are checked against them, on by default.** When a message asks a question about attached documents, and the answer is written from them with no tool call in the turn, a System One verifier reads question, sources and answer. - **What happens
v0.62.3
- **`chimera mcp desktop` now runs with the `mcp` extra alone.** In 0.62.2 it failed on its first tool listing with `No module named 'fastapi'`, because it read the discovery file through a module that imports the web stack. The discovery helpers moved to a module that needs only
v0.62.2
- **`chimera mcp desktop`: Claude can operate the running desktop app.** It is a stdio MCP server that bridges to the app's own local API, and every call runs in the original handler. It is set up in two steps: 1. Turn on **Settings → Claude → "Allow Claude to operate this app"**
v0.62.1
- **The owner's block is read last again.** When `todo_write` was registered, which is every terminal and desktop session, the task-list sentence was appended after the owner's instructions. That broke the precedence study 25 set for every situation: owner > situation > project >
v0.62.0
- **The default agent model is now `openrouter/openai/gpt-6-luna`.** A paired bake-off ran it against `deepseek-v4-flash-0731`, `deepseek-v4.1-flash` and `qwen3.7-flash` on SWE-bench django. The run used the Chimera harness, a network wall and a disk sandbox. - luna resolved **78
v0.61.2
- **`chat`, `assist` and the TUI draw the task list the agent keeps.** `todo_write` is registered on the terminal by default, and the tool itself says it exists so the person watching can see where the agent is. The desktop drew that list; the terminal did not, so the model kept
v0.61.1
- **A reopened turn shows what it was actually charged, beside the estimate.** The receipt's `usd` is priced from the catalogue row of the model id, but one id is served by several routes at different prices. The router's generation record carries `total_cost`, the real charge. T
v0.61.0
- **A typed decision is declared once, and it may only add scrutiny.** Contract v2 (`chimera/decisions/spec.py`): every decision point is a `DecisionSpec` that names its questions, its threshold, its mode (`shadow` by default), what no signal becomes, and the bench that measured
0.60.0 — it continues by itself at the step ceiling, the approval card shows the number that asked, and the sidecar can fetch its browser
- **The Code screen continues a turn that stopped at the step ceiling, by itself, up to three times.** "Continuar sozinho." `max_steps` is the one stop that means *the task was going fine and ran out of room* — the model was working, the ceiling cut it, and the work is incomplete
0.59.0 — hands-free voice that talks while it works, a conversation you can share, and the agent's browser on screen
The nine-item list audited on 2026-09-16 — what the marketing said against what the product did — shipped one pull request per item over two days, each measured before it was built and tested against the running app afterwards. This release is that list, plus what its live tests
0.58.0 — the turn receipt names the route, and a policy review on the Code screen asks you with a card
Two things the live test of 0.57.0 found by using the installed app, fixed the same day.
0.57.0 — what you said at the start survives compaction, and a patch that adds a dangerous call waits for you
When a Code conversation grows past its budget, the app compacts the older part to free room. Until now the dropped span was replaced by a count — *"21 earlier messages were removed"* — which is honest about how much went and says nothing about what still binds. A convention you
0.56.0 — your projects belong to the app, not to one browser window
The Code sidebar groups your conversations by project, and you can add a project before you have ever worked in it. But that list — and the name you give each one — lived in the desktop's own web storage. So it was never really a list of *your* projects: it was a list belonging t
0.55.0 — the approval card asks the safe way, and a turn can stop on its plan
In the desktop app, **"Allow this once" was drawn first and styled as the primary button** — and because the card mounts inside a dialog whose focus trap takes the first focusable element, the caret landed on it. Answering with Enter, without reading, approved. The terminal UI an
0.54.0 — the auditor reads what a slice can do, a runaway is bounded, and the approval question names the command
Two days of measured changes (#425–#447), every one pre-registered, sabotage-verified and published with the predictions that failed.
0.53.0 — the terminal is governed too, and a question finds you
A release about the surfaces a person actually sits at — the terminal, the full-screen app, the desktop chat — which were the ones nothing governed, and about the question that now finds you wherever you are.
0.52.0 — the verify command was the one shell nothing watched
A release about things this project was saying that were not true, and the guards that will not let them come back.
0.51.0 — a tool call now waits for you
The agent used to lose a tool in the middle of a run and not tell you. This release turns that into a question.
0.50.0 — the agent tells you what it is doing, and can reach you
Five things in this release were happening without you being able to see them, reach them, or turn them on.