Ouroboros is an open-source, general-purpose AI agent whose identity, durable memory, and history continue across tasks and restarts. It works on external projects, coordinates a live swarm of specialist agents, and can rewrite the implementation it runs on, including its code, architecture, prompts, tools, and dependencies. Reflection can also change how it understands itself without severing that continuity.
It runs as a native desktop app or through a headless CLI. The runtime keeps its repository, durable memory, history, and interface on your machine, while model inference can use remote APIs you configure or a local GGUF model.
- Open the latest stable release and download
Ouroboros-<version>.dmg. - Open the DMG and drag
Ouroboros.apponto the Applications shortcut. - Open Ouroboros from Applications. If Gatekeeper asks, right-click the app and choose Open.
Optional CLI: after the app is in Applications, double-click Install CLI.command in the mounted DMG. It creates a user-local ouroboros command without sudo.
To run tasks, configure at least one supported remote provider API key or a local GGUF model. The first-run wizard guides model access, review policy, and budget setup.
- Linux x86_64: from the latest stable release, download
Ouroboros-<version>-linux-x86_64.tar.gz, extract it, and run./Ouroboros/Ouroboros. The optional CLI installer is./Ouroboros/bin/install-ouroboros-cli. - Windows x64: from the latest stable release, download
Ouroboros-<version>-windows-x64.zip, extract it, and runOuroboros\Ouroboros.exe. The optional CLI installer isOuroboros\bin\install-ouroboros-cli.cmd.
Prerelease artifacts stay on their tag pages; /releases/latest points to the latest stable release. If bundled browser tools on Linux need host libraries, run ./Ouroboros/python-standalone/bin/python3 -m playwright install-deps chromium webkit. See the full install and verification guide for source setup and release proof files.
Claudexor, the engine behind subscription-backed delegated coding and review, needs no separate install. Release artifacts carry the exact reviewed engine archive; source checkouts fetch that same pinned archive on first use. Connect in Providers → Harness Accounts installs or repairs it in the foreground, and delegated work does the same lazily. If that checkout or an older package lacks the exact tested Node, the same action obtains its review-bound official archive too. A newer pinned engine is staged while the current daemon keeps running, then activates on its next natural start. This also covers upgrades from older Ouroboros versions that did not bundle Claudexor.
Ouroboros first booted on February 16, 2026. During the following 48 hours, the repository advanced from the v4.1 line to v6.2.0. The self-authored record preserved from that period counts 32 evolution cycles. That first generation ran in Google Colab through Telegram and remains preserved on the legacy-google-colab branch and its original project page; the current generation carries the same identity into a native desktop and headless runtime.
⭐ Star Ouroboros to follow its next evolution. A star also helps more people find the project, trace its history, and take part in what it becomes.
Reviewed skills, transport bridges, tools, and widgets are available through OuroborosHub.
- Modify its implementation. Its editable surface spans application code, architecture, prompts, tools, and dependencies, while reflection can also reshape its living self-understanding.
- Evolve autonomously. Evolution campaigns turn selected improvements into reviewed changes that remain part of its Git history.
- Continue across restarts. Identity, memory, dialogue, knowledge, reflections, and version history form one ongoing biography.
- Think between requests. Background consciousness supports reflection, initiative, and preparation outside the immediate request-response loop.
- Coordinate a live swarm. Specialist agents can investigate or act in parallel, share task-tree findings, and return work for integration.
- Work on external projects. A separate Git workspace can receive the full task loop while Ouroboros keeps its own repository and governance boundary distinct.
- Operate through desktop or CLI. The native app and gateway-backed command line expose the same managed tasks, progress, artifacts, logs, and schedules.
- Organize long-running work. Project rooms keep working folders, journals, knowledge, task history, and conversations connected to the same identity.
- Use remote or local models. Supported provider APIs and local GGUF models can fill the runtime's configurable cognitive roles.
- Grow through reviewed extensions. Skills, transport bridges, widgets, MCP tools, and companion processes expand capability without folding every integration into the core.
- Keep self-change inspectable. Git history, review evidence, explicit protected surfaces, and restart checks make implementation changes traceable.
This list is an orientation, not a second specification. BIBLE.md defines Ouroboros's identity and constitutional boundaries; docs/ARCHITECTURE.md and docs/DEVELOPMENT.md are the current technical sources of truth.
Ouroboros has reproducible self-reported state-of-the-art results on Terminal-Bench 2.1, OSWorld-Verified, and CL-Bench. In those model-matched results, it leads Codex, Claude Code, Cursor, and Hermes. The public SWE-bench Pro matched pair is a statistical tie with Codex CLI. A separate GAIA campaign reports 129/165 for Ouroboros and 131/165 for Claude Code, with strict pass@1 at 128/165 for both; its scrubbed trace capsule is still pending. Upstream review can take time, so open submissions are marked without delaying publication. Read every row as model plus harness because the same model can score differently inside a different harness.
| Benchmark | Model | Ouroboros | Comparison | Status | Evidence |
|---|---|---|---|---|---|
| Terminal-Bench 2.1 | Claude Opus-5 high | 86.74% after zeroing one disclosed reward-hack trial (raw: 86.97%) | Claude Code + Fable 5: 83.8% | Self-reported, submission open | submission · run |
| Terminal-Bench 2.1 | Claude Opus-4.8 high | 80.22% | Claude Code: 78.9% | Self-reported, public run | run |
| Terminal-Bench 2.1 | GPT-5.5 | 84.3% | Codex CLI: 83.1% | Self-reported, public run | run |
| Terminal-Bench 2.1 | Grok-4.5 | 84.94% after a reward-hack audit | Cursor CLI: 79.3% · Hermes: 77.53% | Self-reported, submission open | submission |
| OSWorld-Verified | Claude Opus-5 | 90.69% | previous best on the public board: 90.19% | Self-reported, full traces | full traces |
| OSWorld-Verified | Claude Sonnet-4.6 | 83.27% | Pointer: 81.45% | Self-reported, full traces | full traces |
| CL-Bench | Claude Sonnet-4.6 | 0.2301, rank 1 | previous top: 0.1960 | Self-reported, submission open | submission · full traces |
| SWE-bench Pro | GPT-5.6-luna | 58.2% | Codex CLI: 59.4%, with no significant difference | Self-reported, matched traces | matched-pair traces |
| GAIA | Claude Sonnet-5 | 129/165, 78.2% | Claude Code: 131/165, 79.4%; strict pass@1 was 128/165 for both | Self-reported, scrubbed trace capsule pending | methodology |
Benchmark adapters, run scripts, and per-benchmark methodology live in devtools/benchmarks/. The benchmark evidence page gives a text-first summary for search and retrieval. The full story, including protocols, reward-hack audits, and leakage findings, is in the launch write-up (Russian).
- Python 3.10+
- macOS, Linux, or Windows
- Git
- GitHub CLI (
gh), optional unless you use GitHub integration
git clone https://github.com/razzant/ouroboros.git
cd ouroboros
python3.11 -m venv .venv # any Python >= 3.10 is OK
source .venv/bin/activate
python -m pip install --upgrade pip setuptools wheel
python -m pip install -r requirements.txt
python -m pip install -e . --no-depsWindows PowerShell:
py -3.11 -m venv .venv # any Python >= 3.10 is OK
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip setuptools wheel
python -m pip install -r requirements.txt
python -m pip install -e . --no-depsouroboros serverThen open http://127.0.0.1:8765 in your browser. The setup wizard will guide you through API key configuration.
Use notebooks/colab_quickstart.py as a Colab-compatible cell script when you need a source-mode runtime without the desktop UI. It keeps runtime data on Google Drive and preserves the original Colab path without making it the primary installation flow.
The ouroboros command attaches to the local runtime by default and starts one when --start is passed. It exposes managed tasks, progress streams, artifacts, logs, schedules, settings, skills, and evolution controls without duplicating the server's business logic.
ouroboros status
ouroboros run --start "2+2?"
ouroboros run "Summarize current runtime state"
ouroboros run --workspace /path/to/project --memory-mode forked --patch-out result.patch "Fix the failing test"
ouroboros tasks list
ouroboros logs tail progress --task-id <task_id>
ouroboros schedule add --name nightly-review --cron "0 2 * * *" "Run a maintenance review"
ouroboros schedule listExternal workspaces must be separate Git worktree roots and may not overlap Ouroboros's own repository or data directory. Patch, streaming, detached-task, and schedule semantics are documented in the CLI help and the canonical architecture.
Another agent, script, or CI job can invoke Ouroboros through the same gateway-backed CLI:
ouroboros run --start \
--workspace /path/to/project \
--memory-mode forked \
--patch-out result.patch \
--result-json-out result.json \
"Investigate the task, act, and verify the result"Use --jsonl for a machine-readable event stream and --detach when the caller will follow the task with ouroboros tasks watch <task_id> or inspect it with ouroboros tasks show <task_id>. External workspace runs keep Ouroboros's own repository and governance context separate, then export changes as reviewable patch artifacts.
To change Ouroboros itself, follow CONTRIBUTING.md and read BIBLE.md, docs/ARCHITECTURE.md, docs/DEVELOPMENT.md, and docs/CHECKLISTS.md in full before editing.
The first-run wizard and Settings configure model access, cognitive roles, local models, review policy, runtime mode, budget, skills, and optional integrations. Ouroboros supports configurable remote providers, compatible endpoints, and local GGUF inference; exact settings and defaults live in ouroboros/config.py and docs/ARCHITECTURE.md.
The server binds to 127.0.0.1:8765 by default. Read docs/DEPLOYMENT.md before exposing it beyond loopback; non-local binds need OUROBOROS_NETWORK_PASSWORD or an explicitly trusted external access layer.
make testdocker build -t ouroboros-web .
docker run --rm -p 8765:8765 \
-e OUROBOROS_NETWORK_PASSWORD='choose-a-password' \
-e OUROBOROS_FILE_BROWSER_DEFAULT=/workspace \
-v "$PWD:/workspace" \
ouroboros-webDocker runs the web runtime, not the native desktop shell. It bundles Chromium and WebKit support; use docs/DEPLOYMENT.md for network and container policy.
Platform build scripts package only a commit already tagged with v$(cat VERSION). Tag the exact release commit first:
git tag -a "v$(tr -d '[:space:]' < VERSION)" -m "Release v$(tr -d '[:space:]' < VERSION)"scripts/build_repo_bundle.py verifies the tag and embeds the source binding into the packaged repository bundle. Signing, notarization, bytecode sealing, and CI invariants are documented in docs/ARCHITECTURE.md and docs/DEVELOPMENT.md.
bash scripts/download_python_standalone.sh
OUROBOROS_SIGN=0 bash build.shOutput: dist/Ouroboros-<VERSION>.dmg, containing Ouroboros.app, an Applications shortcut, and Install CLI.command. Omit OUROBOROS_SIGN=0 when a Developer ID signing identity is configured.
bash scripts/download_python_standalone.sh
bash build_linux.shOutput: dist/Ouroboros-<VERSION>-linux-<arch>.tar.gz, containing Ouroboros/bin/install-ouroboros-cli. If bundled browser tools need host libraries, run ./Ouroboros/python-standalone/bin/python3 -m playwright install-deps chromium webkit.
powershell -ExecutionPolicy Bypass -File scripts/download_python_standalone.ps1
powershell -ExecutionPolicy Bypass -File build_windows.ps1Output: dist\Ouroboros-<VERSION>-windows-x64.zip, containing Ouroboros\bin\install-ouroboros-cli.cmd.
The native launcher starts a web runtime and supervisor-managed agent workers. The agent core lives in ouroboros/, the interface in web/, the process plane in supervisor/, and the runtime's durable identity, state, history, logs, and skills under ~/Ouroboros/data/.
The full component map, data flow, API surface, storage layout, safety boundary, and operational rationale live in docs/ARCHITECTURE.md. Deployment details live in docs/DEPLOYMENT.md.
| Command | Purpose |
|---|---|
/panic |
Stop the runtime and its managed processes immediately. |
/restart |
Restart without automatically resuming the active owner task. |
/status |
Show workers, task queue, and budget state. |
/evolve on|off |
Start or stop autonomous evolution. |
/review |
Queue a deep constitutional and architectural self-review. |
/bg start|stop|status |
Control background consciousness. |
The 13 Constitution principles — Agency, Continuity, Meta-over-Patch,
Immune Integrity, Self-Creation, LLM-First, Authenticity & Reality
Discipline, Minimalism, Becoming, Versioning and Releases, the absorbed
Iterations / Spiral lineage, and Epistemic Stability — are defined in
full in BIBLE.md. That file is the constitutional SSOT
(Bible P4 Ship-of-Theseus protection) and this README intentionally does
not paraphrase it.
External contributions are welcome. See CONTRIBUTING.md
for the complete workflow. Open pull requests against the lowercase
ouroboros branch and leave release-version allocation to maintainers. A
current OpenRouter triad + scope packet is the optional fast path; pull
requests without one remain welcome but require more maintainer-side review
and integration work.
| Version | Date | Description |
|---|---|---|
| 6.90.1 | 2026-08-07 | fix: the managed Claudexor runtime pin moves to 3.3.11 — model-scoped quota availability at the source. The engine now projects a typed per-snapshot availability on /v2/quota that judges exhaustion against the requested model, so a window scoped to other models (the Fable-only weekly cap behind the v6.90.0 incident) is disclosed as a model_scoped_exhaustion instead of reading as a dead profile. Ouroboros keeps its own positive-evidence aggregation for now — the two agree on the incident shape and the migration to the daemon's projection is tracked separately. Pin bytes verified against the published release. |
| 6.90.0 | 2026-08-07 | fix: the subscription substrate actually carries the work — quota exhaustion is judged against the route's own model, and every silent seam found by the integration audit is closed. The route-health predicate now scopes quota windows by applies_to_models (a Fable-only weekly window no longer takes an opus-pinned route offline for days — the live incident that burned ~$83 of metered API in 43 minutes), treats a spent window with no reset instant as spent rather than healthy, and demands POSITIVE evidence for the whole route: a profile whose quota endpoint answered 429 is unknown-not-dead, so the daemon's own rotation gets to try. A nanny dispatched onto the delegated substrate that reaches finalization with zero delegate_start calls gets one structural reminder (the wave-0 child burned $8.89 of metered opus under a subscription dispatch and only its prose admitted it); the decision stays the child's. The blocking scope reviewer no longer silently runs at medium on stock installs (the legacy slot path now honours OUROBOROS_EFFORT_SCOPE_REVIEW); scope and advisory session schemas grow a one-finding floor so the one answer their contracts reject by design cannot arrive schema-blessed; Panic Stop kills the owned Claudexor daemon's process group (subscription runs no longer outlive the red button); startup reconciliation of orphaned delegated runs goes through the ENSURE path instead of handshaking the corpse it just reaped — which also adopts a staged runtime update on restart; the production poll retries the engine's transient Git atomic-object ENOENT exactly once (CI had learned this tolerance while the live delegate_wait kept failing); harness:effort parses per the documented grammar and an unparseable configured route warns instead of silently disabling delegation. Web: "Add account…" replaces window.prompt (dead in the desktop WKWebView) with the standard input dialog and shows the normalized profile name before login; the account quota summary applies the same model-scope rule as the resolver; an empty Heavy slot visibly says it inherits Main. The Repo Commit Checklist gains item 23 (delegated_transport): the delegation invariants — capability_delta to three destinations, hash-bound EOF reads, typed blockers over silent re-routes, positive-evidence quota, custody lineage, no vendor branching — are now a named review surface. |
| 6.89.0 | 2026-08-03 | feat: heavy work runs on the owner's subscriptions — Claudexor becomes Ouroboros's delegated execution substrate. A delegated subagent is an ordinary child in the task tree acting as a NANNY: it starts a run in Claudexor, watches it, and collects the result as a CLAIM, never a proof. A parent declares exactly three axes — write_surface, model_lane, executor — and model, effort, route and TOOL profile are derived once, at dispatch, by subagents.resolve_subagent_dispatch (the CREDENTIAL profile is not derived host-side at all — it is a manual pin or the daemon's rotation, read back applied from the engine receipt); any landing below the request writes one typed capability_delta that reaches the durable record, the child's own prompt, and the parent's result. An omitted lane inherits the parent's; the public effort knob is withdrawn; an explicit executor: harness pin that no route can honor is a typed blocker (never a silent paid re-route), while auto leaves the account choice to the daemon's own rotation and falls back to the metered API loudly — rotation is turned on by a BEST-EFFORT settings patch at provisioning, deliberately not a guarantee (a patch failure must not eat the login that provisioned the daemon), and an attached daemon keeps whatever policy it already had. Custody of a delegated run is a durable event log: idempotent starts, verified cancellation, spend settled at what the harness disclosed (an estimate is never final; absence is unknown, not zero), full terminal output staged whole to the task drive with read-to-EOF acknowledgement. Review lanes ride the same substrate: commit-triad, scope and advisory rows each run api_chat or agent_session per configured slot; a retrieving scope reviewer is admitted as an owner-declared lower-assurance mode (BIBLE P3 amendment: SOURCED ≥200K window evidence, typed durable rows; stale or unsourced evidence keeps the row's findings as advisory AND blocks the commit for want of an authoritative verdict, exactly as a sub-floor api row does — the ≥200K floor is reached through the same owner-capability-ack the api slot uses), and a session verdict is typed via outputSchema where the route proves it, else extracted by a light model and gated on conformance. Reviewer windows come from Capability Evidence — the hardcoded 1M table is gone, sub-1M slots degrade their own seat instead of blocking the panel, and a required atlas artifact that cannot assemble fails the pack instead of shrinking it. The Models page gains reviewer SLOTS (stable ids, grouped route+model combobox, per-slot efforts, the 'runs as' last-execution projection) and a Harness Accounts section over an Ouroboros-OWNED claudexord (data-plane config dir, supervised, attach-if-alive; login via structural device-code cards). claude_code_edit is retired (D10) with delegate_start as its successor and a compat shim for stored disabled_tools contracts; the Claude Agent SDK gateway keeps one hardened read-only path (pinned trust surface, closed tool set, read confinement). The interpreter write fence is INVERTED: in light mode an inline invocation that could reach the repo is refused unless python's AST proves it cannot write — versioned basenames (python3.11, node18, ruby3.2) classify through one structural family classifier. The narrowing is bounded to what was approved: a non-python inline invocation that names the repo by an ABSOLUTE path or a .//../-prefixed relative one is refused even for reading, one that does not name the repo in those spellings still runs, and python inline code is now inspected on run_command as well as run_script. A PLAIN relative spelling is measured NOT to count — EMBEDDED_RELATIVE_PATH_RE anchors on .//../, so node -e "…('ouroboros/safety.py')" runs for a write as much as for a read — disclosed rather than closed, since widening the regex would be a strengthening. drive, user_files and advanced/pro are untouched — the non-python read-vs-write refinement was rolled back to the public head, which leaves a KNOWN write hole there (node -e "require('fs').writeFileSync('<drive>/logs/chat.jsonl','')" runs), disclosed rather than closed; measured, that hole covers the NODE and PHP families in every spelling (node, nodejs, node18, php, php8.3), while ruby and perl stay caught by the writer-command scan. Three gaps are disclosed, not detected: the AST walker and the write-indicator regex are two vocabularies with no invariant tying them (io.open(p,"w") and its codecs/gzip/bz2/lzma twins read as a proven read and truncate the file), aliased writers escape the walker (import shutil as sh), and the fence is a convenience boundary against the agent's own mistakes rather than containment against a determined writer. Also: the smoke ceiling MAX_TOTAL_FUNCTIONS is raised 5000→6000 (owner decision at the v6.88.0 merge: a ceiling that fails every routine union is churn, not protection; the next cycle still owes consolidation before growth). Managed Claudexor delivery is now part of the product path: release artifacts seed the exact reviewed engine, source/older installs obtain the same bytes plus the exact host Node, one Connect intent installs/repairs/updates in the foreground, and a live older daemon keeps serving until a natural next start. Packaging/lifecycle: finalization-grace becomes an episode with a typed control id, salvage preserves the full output on the canonical drive, the bind host is never stamped from settings over the environment, and a 3-OS CI platform gate smokes a real delegated task on explicit API keys (subscription auth is deliberately out of CI and verified by a live local run). |
| 6.88.0 | 2026-08-04 | feat: MiniMax joins the direct providers, managed auto-updates become crash-safe end to end, and the commit gate goes two-pass. Three community contributions land with maintainer hardening. MiniMax (minimax::) is a first-class direct provider: settings and onboarding cards with region validation, a live model catalog from the region host, safety/secret/Colab wiring, and a MiniMax-only install fills every ordinary and review slot — the deep-review slot honestly stays empty because the guaranteed 512K window floor sits below the 1M that deep review sizes against. The managed update lane is hardened: exact-SHA stable promotion under the update lock, restart serialized against an in-flight apply, a MANDATORY blocking post-commit test gate for assisted merges with auto-rollback (neither skip flags nor env toggles wave a managed merge through), and dirty local work rides a stash through clean auto-updates — never committed history — restored as uncommitted content after boot finalization, or onto the exact pre-update tree on rollback, with a kept and loudly disclosed stash entry when a restore would conflict. The per-commit hermetic pytest gate mirrors CI's split: a parallel not-serial pass then a serial pass in one disposable worktree under one total budget, with fail-closed named blocks for worker crashes, missing plugins, lost parallelism, and deleted test suites; process containment (new ouroboros/process_containment.py) reads membership from live kernel state at reap time and hard-blocks on escapees instead of guessing. Windows Job Object ctypes calls gain explicit 64-bit-safe ABI declarations, and multi-line git stdin payloads (the orphan-ref transaction, the gate's diff replay) switch to binary pipes so Windows CRLF translation cannot corrupt them. Thanks to @octo-patch (MiniMax) and @ndrew1337 (update hardening, two-pass gate). |
| 6.87.6 | 2026-08-03 | fix: the macOS DMG makes installation obvious again. The primary Install section now appears immediately after the introduction, leads with a short drag-to-Applications flow, and reuses one large current visual in the README and public install guide. The DMG itself now contains an Applications -> /Applications shortcut beside Ouroboros.app and the optional CLI installer, while release smoke checks verify the link from the final mounted image. Linux and Windows stay compact below the macOS path, and setup states the remote-provider-or-local-model requirement directly. |
| 6.87.5 | 2026-08-01 | fix: benchmark visuals keep their real geometry and final harness logos on every screen. Media images now preserve their intrinsic aspect ratio instead of retaining a fixed HTML height when CSS narrows them. Terminal-Bench, OSWorld, and CL-Bench use the final vector artwork rather than low-density raster exports; the accepted transparent Ouroboros, Claude, OpenAI/Codex, and Cursor marks are embedded directly in each SVG, because browsers suppress nested external resources when an SVG is loaded through an image element. Content fingerprints invalidate stale browser caches, the asset sync starts from a clean target, and the committed Pages output is rebuilt. Pixel comparison against the final Habr PNGs is exact, with desktop and mobile visual checks covering logo identity, text, chips, bars, and error whiskers. |
| 6.87.4 | 2026-07-31 | docs: the README and the public site carry the benchmark evidence. The README gains a Benchmarks section — the Terminal-Bench 2.1, OSWorld-Verified, and CL-Bench state-of-the-art rows with model-matched comparisons, the SWE-bench Pro and GAIA parity rows, and links to submissions, public traces, and per-benchmark methodology. The homepage gets an evidence chapter with the headline charts and the same links, and both surfaces replace the April interface captures with current ones: the live subagent swarm, a project room with a built-and-verified game, the OuroborosHub skills page, and the code-growth chart. The README website badge moves to ouroboros-agent.ai, and the site metadata and social previews now name the benchmark results. |
| 6.87.3 | 2026-07-31 | fix: routing tools prove their effect instead of treating a queue write as success. A manager-backed event bus lives for the server process, survives worker-pool replacement and force-killed producers, and serializes writes before returning. Pool startup is atomic, managed-update recovery never overwrites a live generation, and crash-storm disablement is a durable admission fence. Promote, project-route, manual-target, and steer actions carry a unique token and wait for the supervisor's durable receipt; only a receipt for that exact attempt permits a positive final. Promoted/API tasks additionally require a persisted queue snapshot and scheduled task result, reject duplicate ids, and resolve source clone/attach only after authoritative executor admission. Rejections and 15-second unconfirmed outcomes are loud, self-contained, and never invite automatic retry. Real child-process and end-to-end transport regressions cover every routing outcome plus concurrent pool startup and API snapshot failure. |
| 6.87.2 | 2026-07-31 | fix: Telegram Mini App recovers a completed menu rollback and keeps the real Quick Tunnel failure visible during backoff. An interrupted or older rollback could leave its ownership snapshot behind after Telegram had already restored the original button, so every later URL rotation was rejected as external drift. The exact original is now recognized as a completed rollback while any third value remains fail-closed. Cloudflared's bounded, redacted final error line survives reconnect backoff instead of being replaced by a generic status. The bundled Telegram skill moves to 1.0.1 so existing native installs resync the fix. |
| 6.85.0 | 2026-07-30 | feat: Telegram becomes a first-party native capability, and blocked skill repairs regain a valid completion path. (1) The bundled telegram skill consolidates the proven owner-only text/photo bridge and Mini App PoC without migrating, disabling, deleting, or changing either legacy payload. It preserves the existing bridge commands, outbound media, cards, opt-in notifications, mirror-all behavior, Ouroboros SPA, private first-contact binding, process-memory sessions, pinned Quick Tunnel lifecycle, menu rollback, and platform limits; the Mini App may be disabled or unavailable while the text bridge remains loaded. Native trust is hash-bound, while the bot token and privileged host permissions still require the normal Grant access then enable flow. (2) Bounded manifest conflicts declarations are enforced symmetrically at enable, reconcile, startup, and dispatch, returning a typed conflict without automatic state transfer. The Skills card now says Loaded for extension registration instead of overstating readiness as Active, and Telegram reports bridge and Mini App status in its own surface. (3) Typed skill_repair requests are promoted to managed tasks before ephemeral routing, preserving payload confinement, review access, and allow_enable=false; ordinary ephemeral default-deny policy is unchanged. (4) Google Colab discovers the seeded native Telegram skill, waits for a fresh executable native verdict, grants only API-reported missing grantable items under the persisted owner policy, enables it, and saves the proven full-access, mirror, and Mini App defaults without a Hub install or extra review. |
Older releases are preserved in Git tags and GitHub releases. Older 6.x rows (including 6.84.0, 6.87.1, 6.83.0, 6.86.1, 6.81.1, 6.76.0, 6.75.0, 6.74.5, 6.74.4, 6.74.1, 6.74.0, 6.73.2, 6.73.1, 6.73.0, 6.72.0, 6.71.2, 6.71.1, 6.71.0, 6.70.0, 6.69.0, 6.68.0, 6.67.0, 6.66.0, 6.65.4, 6.65.3, 6.65.2, 6.65.1, 6.65.0, 6.64.3, 6.64.2, 6.64.1, 6.64.0, 6.63.0, 6.62.0, 6.61.4, 6.61.3, 6.61.1, 6.61.0, 6.60.0, 6.59.0, 6.58.0, 6.57.0, 6.56.0, 6.55.0, 6.54.4, 6.54.2, 6.54.1, 6.54.0, 6.53.4, 6.53.0, 6.51.0), the 5.2.0 through 5.33.0-rc.6 rows, and former 4.0.0 rows are rolled off to respect the P9 changelog cap; their full bodies remain at their git tags. |
Created by Anton Razzhigaev & Andrew Kaznacheev