Releases
Changelog
Two changelogs, one for each product: Chronos DB and Chronos Gateway. Every release, newest first, with everything that changed in it. Nothing is left out: these are the notes each release ships with.
Chronos DB
The engine: one binary, chronos, that you run yourself. 28 releases since 2026-09. All of them on one page, downloads, docs.
Latest: 0.6.2
Strict proposals for payments, safer attached databases, and fixes from an audit of 0.6.0. A proposal that would send a payment, payout, transfer or refund is now strict: only a person approves it, within 15 minutes, and any commit to main since it was proposed supersedes it. Beside Postgres or MySQL, merges into main are refused when the source may lose a commit it acknowledged; beside MongoDB, merges commit to a majority, journaled, and a merge whose outcome was lost is no longer dropped when it in fact committed. chronos refresh --check compares main with the source and changes nothing. Nothing changes in how rows are stored: 0.6.2 opens and writes data format 8 as 0.6.0 does. Built and signed on the maintainer's Mac as 0.6.1 was.
- Attached databases: merges into main are refused when the source may lose a commit it acknowledged (Postgres with
fsyncorsynchronous_commitoff; MySQL withinnodb_flush_log_at_trx_commitnot 1, orsync_binlognot 1 with the binary log on), nothing written, saying which setting.CHRONOS_ATTACH_UNSAFE_SOURCE=1merges anyway. - Attached MongoDB: a merge's commit is acknowledged by a majority of the replica set, journaled, and its reads use read concern
majority. The probe that settles a merge whose commit outcome was lost now runs in a transaction it always aborts. The old probe took a merge still committing for one that never would, and dropped it, so a merge that then committed in MongoDB never reached main; it also could leave a probe document behind. A probe that can't tell yet (the merge's transaction still open) keeps the merge in doubt, and merges into main wait. chronos refresh <folder> --check: the reconciliation scan. Compares main with Postgres as a refresh does and takes nothing: per table, rows only in Postgres, rows that differ, rows only in main, and tables whose columns differ. Exit 1 if they differ. Postgres sources only for now.- Fixes from an audit of 0.6.0.
- A Postgres client that stops reading (a big result it never takes) no longer blocks its session, and keeps its locks, forever: a write that waits a minute ends the connection. TCP keepalive finds clients gone without a word.
- One address can no longer fill every connection slot with connections that never log in: at most 64 from it may be logging in at once; the rest are refused as "too many clients".
- A new agent can't take the name of a dropped one while grants on worlds still name it: it would have had them. An agent a merge policy names as its reviewer can't be dropped until the policy changes.
- An agent's
max_worldsholds against forks sent at once (each counted before any landed, and all went through). SHOW WORLDS(and the HTTPbranchesop) to an agent held to a tenant lists main and its tenant's worlds only.- A merge retried by its operation id while the first try is still merging returns the first's result, not "no such world".
- While the checkpoint that upgrades a folder to format 8 fails (a full disk, a store down), merges into main are refused instead of landing with no ledger row; each merge tries the upgrade again.
- With a transaction left open, commits no longer slow down as row locks pile up: the list is pruned again only once it doubles.
- Statistics of dropped, merged and discarded worlds are freed, as a process running millions of short task worlds kept them all.
- Attached databases:
chronos attachandchronos servesay at the start when the source may lose a commit it acknowledged, and so refuses merges into main, instead of leaving it to the first merge. - Attached databases: a merge the source committed before a crash cut it short is made in Chronos as it was decided. Recovery no longer judges it again by a merge policy, merge checks, scope or stale reads, which could refuse it and leave the world open with its rows already in the source; nor as the agent whose merge happened to start the recovery. Its effects keep their delay, and its rows the taint the merge gave them. A merge recorded by 0.6.0, which doesn't name the world's version it merged at, is judged as before.
chronos refresh <folder> --checkworks beside MySQL, MariaDB and MongoDB too, comparing main with the source as their refreshes do.ASSIGN WORLDwrites the whole tree's tenant in one change: anALTER WORLD, a discard or a fork of one of its worlds can no longer land between its checks and its writes and leave the root assigned with a descendant not. A world forked from the tree meanwhile refuses the assignment (40001, try again), and a fork that started before the assignment landed refuses itself instead of staying unassigned under an assigned world. The same statement (tenant and reason) run again after a crash cut one short finishes the worlds left.CONSOLIDATE: its scratch world (_consolidate_<p>) can't be dropped while the consolidation runs. A drop between marking the children and the scratch world's merge into p, then a crash, discarded children whose rows never reached p.CONSOLIDATEgiven a world's id works as given its name (it found none of the world's children).- Strict proposals for payments (rules E7 and E8). A proposal whose world queues an effect on a payment-like channel (
payment*,payout*,transfer*,refund*, or a policy'sreview_channelsor newstrict_channels) is strict: only a person approves it, within the policy's newstrict_window(15 minutes unless set; then it'sexpired), and any commit to the world it merges into since it was proposed supersedes it, at approval or at the merge (40001,strict_target), so a payment never goes out against a state nobody approved. It merges whole into its parent.SHOW PROPOSALSshowsstrictandstrict_until;SHOW MERGE POLICIESthe two settings;EXPLAIN MERGEastrictline before anything is proposed.
All 28 releases
- 0.6.2Strict proposals for payments, safer attached databases, and fixes from an audit of 0.6.019 November 2026 · beta · 10 changes
- 0.6.1Two fixes for merges cut short by a crash18 November 2026 · beta · 2 changes
- 0.6.0Merges are logged as what they did, agents are kept to their tenant, and worlds gain what long-running agents need18 November 2026 · beta · 17 changes
- 0.5.5Deeply nested statements run again, and nested calls no longer take minutes17 November 2026 · beta · 1 change
- 0.5.4Faster scans, `GROUP BY` and one-key joins16 November 2026 · beta · 1 change
- 0.5.3Faster undo of big merges, and faster analytical joins16 November 2026 · beta · 2 changes
- 0.5.2A merge undone in part can be undone again16 November 2026 · beta · 4 changes
- 0.5.1Follow MongoDB's changes as they happen, and a fix for least-privilege MySQL users16 November 2026 · beta · 3 changes
- 0.5.0Attach beside MySQL, MariaDB and MongoDB, undo a merge by its id, and five security fixes15 November 2026 · beta · 16 changes
- 0.4.10Rules for agents that run on their own10 November 2026 · beta · 17 changes
- 0.4.9Writes that fit the database you already run9 November 2026 · beta · 8 changes
- 0.4.8Counts read right, and rustls with its security fixes3 November 2026 · beta · 2 changes
- 0.4.7The Studio, in the look of chronos.2090.tech3 November 2026 · beta · 2 changes
- 0.4.6A tenant's run is held only by its own tenant's rows1 November 2026 · beta · 1 change
- 0.4.5A run held to a tenant can change child tables31 October 2026 · beta · 1 change
- 0.4.4Attach to the Postgres you already run: tested on Postgres 12 to 18 and Supabase, and tables with row-level security, without a primary key, with generated columns or partitioned are attached. Main keeps up with Postgres as it changes (`serve --follow`), masked columns work on attached folders, and agents held to a tenant are held to it in child tables too31 October 2026 · beta · 12 changes
- 0.4.3Postgres compatibility from running real apps (PostgREST and Supabase's setup, psycopg's server-side cursors, Django's `explain()`, deferred UNIQUE constraints, a third batch of PostgreSQL 16's regression suite), and agents on production data held tighter: masked columns that triggers can't leak, tenants that must name a real column, low-priority scans, and an attached Postgres's write-back you can rehearse and undo30 October 2026 · beta · 38 changes
- 0.4.2Joins, merges and memory, from running BranchBench (Columbia's benchmark of agent workflows on branching databases): joins on several columns and joins too big for memory, merges where both sides changed a table's columns, a merge that counts its conflicts, `DIFF ... SUMMARY` and `UNDO DIFF`, and a memory limit for every query25 October 2026 · beta · 26 changes
- 0.4.1Fixes from the first run of `chronos init` against Chronos Cloud, attach that keeps up with Postgres, and agents held to one tenant23 October 2026 · beta · 8 changes
- 0.4.0Chronos Cloud from your terminal, and a server that's safe by default22 October 2026 · beta · 9 changes
- 0.3.2Agents can check their reads between steps, and a vetoed effect's record is safer to keep17 October 2026 · beta · 5 changes
- 0.3.1Fixes for importing from Supabase (and any Postgres with row-level security, or extensions in a schema of their own)17 October 2026 · beta · 5 changes
- 0.3.0Reviews you can trust with production data, from the first week of feedback on 0.2.017 October 2026 · beta · 5 changes
- 0.2.0The first public beta: agents' changes to your Postgres data, reviewed and checked before they land17 October 2026 · beta · 518 changes
- 0.1.3Vector search at 1M rows: 2.1× faster at the same recall, 3.5× at the new default beam, a third less memory to build29 September 2026 · 16 changes
- 0.1.2Move from Postgres in one command: `chronos import <folder> postgres://...`28 September 2026 · 24 changes
- 0.1.1Merge policies: agents merge on their own within rules, the rest wait for a person27 September 2026 · 18 changes
- 0.1.0A misspelled column suggests the nearest one; COPY ... FROM STDIN (FORMAT binary); COPY streams into its INSERT27 September 2026 · 192 changes
Chronos Gateway
The gateway that puts agents on a database you already run, with approvals. It ships with the engine it was tested on. 24 releases since 2026-10. All of them on one page, downloads, docs.
Latest: 0.4.4
- The plain-words diff of a run never shows a masked column's value. Its changes (an approval link, the console,
GET /v1/runs/:id/changes) read(masked)there, as the undo preview already did; the engine's admin diff shows them. - A deleted webhook gets nothing more. Deliveries still due to it were sent on through their retries, about 8.6 hours, approval links in Slack messages included; they are dropped now.
- A webhook's URL is never logged, and its secret is kept sealed. A Slack webhook's URL is its credential: the give-up warning names the webhook's id, and the error kept with a delivery leaves the URL out. Secrets made before are read as they are.
- One receiver that doesn't answer no longer holds the others back. A webhook or an effects channel whose receiver timed out (or refused the connection) is tried once a pass; its other deliveries wait for the next, so Slack approvals and refunds on other channels go out meanwhile.
- A Chronos Cloud undo of a run is recorded in the run, as through the API: the run says
undone,merge.undonegoes to the webhooks, and the run's merges are undone together (a part of them, 400). - MongoDB tools write all of a call's documents or none.
mongo_insertandmongo_update_manycheck every document before the first is written: a bad one refused the call after the ones before it were written, and a retry wrote those twice. - MongoDB updates work beside a masked ObjectId or integer field. A lifted field the update leaves alone is neither checked nor written; one the run read masked (as text) failed every update of the collection.
- A MongoDB tool call counts its writes against
GATEWAY_RUN_STATEMENTS_PER_MINUTE, before the first: one call counted once however many documents it wrote. A call with more writes than the limit left is refused whole (429). - A new setup code works after a lockout. The count of wrong setup codes is kept with the code, so one made by
chronos-gateway setup-code(another process) starts again, as the error says; before, only a restart did. deploy/compose.yamlpullslatestby default, not 0.2.11, whose image can't start its engine.- The run list, status counts and sweeper use indexes, and webhook deliveries done with for 30 days are deleted once an hour: each of these read every run, under the store's one lock, at every call.
All 24 releases
- 0.4.4The plain-words diff of a run never shows a masked column's value; A deleted webhook gets nothing more; A webhook's URL is never logged, and its secret is kept sealed19 November 2026 · with Chronos DB 0.6.2 · 11 changes
- 0.4.3A merge from Chronos Cloud is a person's approval of a run, as through the API; Masked and hidden columns no longer reach the cloud in diffs mode by other ways; A late approval that meets a conflict is kept until the hold ends19 November 2026 · with Chronos DB 0.6.2 · 6 changes
- 0.4.2A strict change is superseded by a commit to the database since its hold19 November 2026 · with Chronos DB 0.6.2 · 1 change
- 0.4.1Engine 0.6.2; Strict changes are for a person, in their window19 November 2026 · with Chronos DB 0.6.2 · 2 changes
- 0.4.0Engine 0.6.0; Runs are safe against lost answers and two decisions at once; Child runs18 November 2026 · with Chronos DB 0.6.0 · 7 changes
- 0.3.4Deeply nested SQL runs again, and reads are faster17 November 2026 · with Chronos DB 0.5.5 · 1 change
- 0.3.3Undo of a big run is about 20 times faster16 November 2026 · with Chronos DB 0.5.3 · 1 change
- 0.3.2Undo after a partial undo; Undo after a part was approved16 November 2026 · with Chronos DB 0.5.2 · 7 changes
- 0.3.1GATEWAY_FOLLOW beside MongoDB; MongoDB filters: $regex and dates; Undo from Slack16 November 2026 · with Chronos DB 0.5.1 · 7 changes
- 0.3.0The image can start its engine again; MySQL and MariaDB; Undo a merged run15 November 2026 · with Chronos DB 0.5.0 · 12 changes
- 0.2.11An AI reviewer (any OpenAI-compatible model) decides held merges before a person, seeing structure rather than row text.10 November 2026 · with Chronos DB 0.4.10 · 3 changes
- 0.2.10Runs can write as the end user's Postgres role (act_as), so Postgres's grants and row-level security judge them.9 November 2026 · with Chronos DB 0.4.9 · 1 change
- 0.2.9People allowed by a group can approve; signing out asks the provider again; a refused person can try another account.3 November 2026 · with Chronos DB 0.4.8 · 1 change
- 0.2.8A person holding an approval link no longer sees a conflict's rows, hidden columns included.3 November 2026 · with Chronos DB 0.4.8 · 1 change
- 0.2.7The console, approval widget and approval pages redrawn in Signal; --chronos-on-accent.2 November 2026 · with Chronos DB 0.4.6 · 1 change
- 0.2.6Signed downloads and image (cosign). The installer links to the guide.1 November 2026 · with Chronos DB 0.4.6 · 1 change
- 0.2.5A tenant's run is held only by its own tenant's rows.1 November 2026 · with Chronos DB 0.4.6 · 1 change
- 0.2.4The connection check accepts providers' certificates; a person may approve a run whose reads changed.31 October 2026 · with Chronos DB 0.4.5 · 1 change
- 0.2.3Neon; runs that change child tables.31 October 2026 · with Chronos DB 0.4.5 · 1 change
- 0.2.2Fixes from a clean-machine install.31 October 2026 · with Chronos DB 0.4.4 · 1 change
- 0.2.1Console fixes.31 October 2026 · with Chronos DB 0.4.4 · 1 change
- 0.2.0The admin console.31 October 2026 · with Chronos DB 0.4.4 · 1 change
- 0.1.128 October 2026 · with Chronos DB 0.4.2 · 0 changes
- 0.1.0First release.28 October 2026 · with Chronos DB 0.4.2 · 1 change