Skip to content

AgentRun v0.1.0-beta.41

Pre-release
Pre-release

Choose a tag to compare

@claudio-michel claudio-michel released this 16 Nov 21:06
· 230 commits to main since this release
fa4da5a

AgentRun 0.1.0-beta.41

AgentRun is an open-source platform that supercharges your agents. It adds self-improvement, durable execution and a
control plane on top of your favorite agent harness. Your agents become smarter, more fault tolerant and use fewer
tokens with every run. This release is about agents that watch the world for you. A feed trigger starts a run for each
new or updated item of an RSS, Atom or JSON feed, or a JSON list such as GitHub's pull requests, and polls, so nothing
outside has to reach the server. Settings → MCP servers adds the servers the free templates use in one click: Parallel Search,
Firecrawl, GitHub with a token, Linear and Notion, and a template that needs one of them links straight to it filled. A
Google-backed MCP server stays signed in, a run's learned workflow tool runs only the revision that serves, and a
SQLite server no longer fails a write StoreBusy when its own transaction lets go during the wait.
A run keeps the labels its caller gives it and GET /runs filters on them, the server takes a list of API tokens so one
rotates with no caller refused, and a message to a working claude-agent run steers the turn in flight. On Postgres the
store moves to version 16, so going back to beta.40 there is restoring the copy taken before the upgrade; on SQLite it
keeps version 11 and going back is changing the image back. The launcher is
@parcha/agentrun 0.4.9, the Python client parcha-agentrun is 0.3.9, and the server image is
ghcr.io/parcha-ai/agentrun-server:v0.1.0-beta.41 (linux/amd64 and linux/arm64).

No fresh store: beta.41 opens the data beta.40 left. Keep the same data folder and the same start command. On the
Docker image that is --data /data/b37 --listen 0.0.0.0 --port 7340 --drain-timeout 6s as before, and on Render the
Docker Command you set for beta.37 stays. Change only the image to ghcr.io/parcha-ai/agentrun-server:v0.1.0-beta.41.
The console keeps its login, and your agents, MCP servers, keys, sign-ins, email account and triggers are all still
there. Upgrade the server first, then the runners: a beta.41 runner opening a data folder its server has not opened
yet is refused ("Start the AgentRun server to add each run's labels before opening a runner") and changes nothing.

  • SQLite keeps schema version 11: the server adds the nullable runs.labels column and the minted_tokens table at
    its open, and every run started before reads as given no labels.
  • Postgres goes from version 14 to 16 at the server's start: 0015_run_labels (a nullable runs.labels column) and
    0016_minted_tokens (a new table). With the server and runners stopped, before the first beta.41 start, take
    pg_dump --schema="${AGENTRUN_STORE_SCHEMA:-agentrun}" <url> > dump.sql, in a shell where AGENTRUN_STORE_SCHEMA is
    what the server is given, if anything (the schema is agentrun when it is unset).
  • Going back. On SQLite, set the image back to v0.1.0-beta.40; the command stays. On Postgres there is no way back
    in place: beta.40 refuses a schema at version 16 ("set up by a later release than this one"). Going back is
    psql <url> -c "DROP SCHEMA \"${AGENTRUN_STORE_SCHEMA:-agentrun}\" CASCADE", then psql <url> < dump.sql with
    psql from the same Postgres release as pg_dump, and setting the image back, which drops everything done on
    beta.41 since.
  • Feed triggers and going back. beta.40 turns every feed trigger off at its start (source_unreadable, "this
    server has no feed source"), and it stays off when you return to beta.41: turn each on again (PATCH /v2/triggers/{id} {enabled: true}). It watches from that poll, so items that arrived meanwhile start no run.

Highlights

  • A feed trigger. {kind: "feed", urls, every_s?: 900, items?, id?, accept?} starts a run for each new or updated
    item in 1 to 20 feeds: RSS, Atom, JSON Feed, or a JSON list such as GitHub's pulls API, read at its items and id
    pointers. It starts watching at its first poll, which a create or turn-on runs at once (10 seconds at most), so the
    items already there start no run. An item fires once, and again with update: true when its updated time moves on;
    one poll starts 20 runs at most, oldest first, and reads are conditional on ETag and Last-Modified. A 410, or a 404
    three checks running, stops that URL, and the trigger turns itself off (source_unreadable) only when every URL has
    stopped. Reads are held to the MCP servers' address check (no private, link-local or own address, each of 5 redirects
    checked again), 20 s and 5 MB at most, and a feed that declares an entity in a DTD is refused. The run gets
    {id, link, title, published, updated, author, summary, categories, feed: {title, url}, update}, or
    {id, link, title, item, update} for a JSON list. Template packages may carry feed triggers, with urls from the
    owner's parameters, and a template eval case's feed: fixture serves fake feeds (#3005). The console makes and edits
    one: Triggers → New feed trigger takes 1 to 20 URLs, one a line, an interval, a JSON list's pointers and "Only when"
    conditions, and the trigger's page says its last check and what it starts from ("Watching 2 feeds from now: the 130
    items already there won't start runs.") (#3020).
  • Run labels. A run takes labels at its start (POST /agents/{agent_id}/runs, POST /runs): up to 16 of the
    caller's own key-value strings, served on every read, and GET /runs and GET /agents/{agent_id}/runs take
    label=<key>:<value> to list only the runs holding it. The server reads nothing else from them, so they are no place
    for a secret (#2214).
  • API tokens that rotate. AGENTRUN_TOKENS holds a list of id:secret entries, any of which opens every route the
    token opens, so a token rotates with no caller refused: add the new one, move the callers, remove the old one.
    AGENTRUN_TOKEN (or DIR/.token) keeps working as the one entry default. Each request's log line names the token's
    id, never its secret, and deploy/README.md gives the three steps (#2217).
  • Steer a Claude Agent run mid-turn. A message sent to a working claude-agent run is read at the model's next step,
    in the same turn, where it waited for the turn's end; the Timeline shows it after the tool call it arrived during
    (#2783).
  • Composio deletes clean up after themselves. Deleting a Composio server also removes what its save made in your
    Composio account (the connected account its link started and the auth config it made), keeps what another saved server
    still uses and never touches what was there before, and says what it removed and kept (#3015).
  • Learned workflows re-pick a quote miss. A learned workflow whose answer quotes the pages it fetched writes its
    answer once more on a quote miss before it hands back to the agent; on the test set, re-picking kept 26 of 31 misses at
    about half the cost of handing every one back (#3016).
  • MCP presets. Settings → MCP servers offers five servers in one click, the ones the free templates use: Parallel
    Search (https://mcp.parallel.ai/mcp, no credential), Firecrawl (https://mcp.firecrawl.dev/v2/mcp, free and rate
    limited without a key), GitHub with a fine-grained token (a Read-only switch picks .../mcp/readonly), and Linear and
    Notion with Sign in. Each fills the name, address, credential and toolkit; #/settings/mcp?add=<name>&preset=<id> opens
    the form filled (#2999). Applying a template whose url MCP need is one of these toolkits links to that preset filled
    (#3003). When the keyless web search is rate-limited, its words name the free Parallel Search server too, and the
    Parallel preset uses the address Parallel's docs give (search.parallel.ai/mcp is its legacy one) (#3006).
  • A Google MCP server stays signed in. A sign-in whose authorization endpoint is on accounts.google.com asks for
    offline access (access_type=offline, prompt=consent), so it gets a refresh token; other servers' sign-ins are
    unchanged (#2997).
  • The learned workflow tool runs only what serves. A run's learned_workflow tool calls a workflow's champion, the
    revision promoted last, never one that is replaying, queued, held or undecided. Before, a run its workflow did not serve
    first could be handed the folder's newest revision, so an unpromoted draft could answer it (#3008).
  • SQLite writes that wait land. A begin, write or statement whose wait for the write lock ran out while a transaction
    of the same process held it no longer fails StoreBusy when that transaction's commit is read before the busy answer: it
    waits on and lands (#3002).
  • Computer vendors. Vendors that build images their own way get two more prepareImage paths: a push to the vendor's
    registry tagged agentrun-box:<hash>, or box/Dockerfile as the build context on the vendor's side (#3000). Output that
    comes as server-sent events reaches framedExec with no reader of its own (#2996). Modal's kill() ends the process
    tree in the machine and Daytona's exec runs on framedExec (#3001), the exec pre-check stays under Modal's 64 KiB
    argv cap (#3009), and a vendor's deleted machine reads undefined (#2992). A vendor machine with no exec can dial the
    server instead, with a token minted for it alone, and fetch the server's own computerd build (#3013, #3019). A
    self-hosted Daytona target with an Archil egress relay (DAYTONA_ARCHIL_EGRESS) takes Archil mounts again (#3021).
    packages/computer/SCORECARD.md reads 25 sandbox vendors against the provider contract (#2994).

Breaking

  • The machine-host label and lifetime mechanisms are cut back to what the reference providers use (#2998,
    @parcha/agentrun-computer). HostFacts.backstop, the Core case provider.backstop, HostFacts.idleMs and the host
    member keepalive are gone, and the shared adapter packs labels itself. A host that declares any of these, or packs its
    own labels, breaks:
    • backstop: true: delete it; req.backstopAt stays for a janitor host's delete timer.
    • idleMs and keepalive: delete them and switch off every vendor timer that ends a quiet machine.
    • labels.packed: stop calling packLabels/unpackLabels; send req.labels as they come and filter list by them.
    • labels: "machine": declare lifetime: "vendor", leave extend out, and answer each machine's image from get
      and list as the new Machine.image.
  • MachineHost.exec is optional (#3013, @parcha/agentrun-computer): a host whose facts say connect: "dial" has
    no exec and no upload, so code that calls a host's exec itself checks it is there first. HostedProvider takes
    dial: { url, tokens, hub }.
  • A Composio server's delete answers 200 with a body (#3015): DELETE /v2/settings/mcp-servers/{name} on a
    Composio server answers {deleted: true, composio: {removed, note}} where a full removal answered 204. A save as
    another user_id or auth config while Composio triggers use the server is refused 409 composio_in_use, and a save
    over a server a concurrent Composio save moved is refused 412 precondition_failed.

Behaviour changes

  • A beta.41 runner refuses a data folder its server has not opened yet (#2214); start the server first.
  • POST /v2/runs/{id}/messages answers mode steer for a working claude-agent run, where it answered follow_up,
    and GET /v2/capabilities lists claude-agent with steer: true (#2783).
  • An SDK run's credential (ar_sdk_…) is signed with a secret derived from the seal key, not the API token, so no
    change to the token list strands a live run; one signed by an earlier server works while its token is listed (#2217).
  • The launcher takes the computer vendors' keys out of the server's environment (ARCHIL_API_KEY, E2B_API_KEY,
    ORGO_API_KEY, PHONE_HARNESS_API_KEY, AGENT37_API_KEY, AGENT37_SSH_KEY) and hands them over its pipe, as it did
    Daytona's and Modal's. An RSA AGENT37_SSH_KEY can take the secrets past the pipe's 4,096 bytes, and then the launcher
    refuses to start: save that key in Settings → Computers instead, or use an Ed25519 key (#2777).
  • A learned workflow's next compiled revision must carry the quote re-pick loop to pass the learning checks; one
    already merged keeps serving as it is (#3016).
  • The runner beside a local install refuses an MCP call, a refresh or a pin's listing at the server's own ports,
    as the settings routes do (#3022).
  • Settings → Computers refuses a key saved as whitespace alone (422 wrong_type) (#3021).
  • A dial provider's create refuses without AGENTRUN_PUBLIC_URL, and a runner in its own process has no dial
    (#3019).
  • Model requests AgentRun sends to OpenRouter carry OpenRouter's app attribution headers (HTTP-Referer: https://github.com/Parcha-ai/agentrun, X-OpenRouter-Title: AgentRun, X-OpenRouter-Categories), so usage counts
    toward AgentRun's page on OpenRouter, a self-hosted server's on its owner's own key included. They go only to
    openrouter.ai itself, from every harness and the server's Decisions providers. Set AGENTRUN_OPENROUTER_ATTRIBUTION=off
    where AgentRun starts to send none (#3029).
  • A learned workflow with nothing promoted is left out of a run's learned_workflow tool (#3008); one no promotion
    record names yet (no run's input fit it) is offered as before.
  • A Google MCP sign-in shows Google's consent screen each time (prompt=consent), which is what returns the refresh
    token (#2997).
  • Choosing an MCP preset clears a token, header rows, scopes or client typed for another server, and a new server
    whose name is already saved can't be saved or signed in (#2999).
  • The exec pre-check's kill check now passes on Modal; the Core case provider.prepare-image fails a provider whose
    prepareImage() method and descriptor flag disagree
    (#3001).
  • With labels: "machine", a machine whose label read fails or doesn't end is left out of that listing with its id
    and why through onNote; a create's listing by its name fails instead (#2998).
  • A triggered template eval case, an email one included, puts its mail or item in only once the trigger's first poll
    has read
    , so a slow first poll no longer takes it as part of the start; a template eval's trigger clock fires a timer
    it moves past at once (#3005).
  • A template's feed trigger whose array parameter is given empty is skipped with a triggers.<i> warning, and the
    agent stands (#3005).

Known issues

Known issue, Mode 2 (the run folder on your Archil disk):

  • Works: sandboxes on Agent37, Modal and E2B, and desktops on Agent37 and E2B, in any account. The first computer in an account starts from AgentRun's public box image: Agent37 registers it once ("pulling" for about 15 s), Modal pulls it once (about 35 s), E2B starts from a public template in about 5-10 s. Shell commands start in /workspace, the run folder.
  • Not yet: Modal sandboxes with GPUs (the Archil mount isn't proven on Modal's GPU runtime), and Daytona's hosted cloud (no Archil mount there). A self-hosted Daytona target takes Mode 2 once its operator sets DAYTONA_ARCHIL_EGRESS and AGENTRUN_DAYTONA_SNAPSHOT names a snapshot of box/Dockerfile.
  • What to do: for those two, use another provider for Mode 2, or set the computer to none.

Known issue, spend caps: a cap below what the model needs for one step and an answer wraps up early; saving such an agent now warns.

Dependencies

  • @parcha/agentrun-serve adds fast-xml-parser 4.5.7, pinned exact (with strnum 1.1.2), the feed trigger's parser
    (#3005).
  • The repository root adds marked 18.0.11 as a dev dependency, the version the root lockfile already held for pi-tui,
    for the docs link check; no published package changes with it (#3010).

What you can do now

  • Add Parallel Search, Firecrawl, GitHub (token), Linear or Notion as an MCP server in one click, or link someone to the
    form filled (#2999, #3003, #3006).
  • Keep a Google-backed MCP server signed in past its first hour (#2997).
  • Start a run for each new item of a feed: a blog, a vendor's status page, a changelog, a repository's pull requests,
    made and edited in the console (#3005, #3020).
  • Label runs at their start and list them by label (#2214).
  • Rotate the server's API token with no caller refused (#2217).
  • Steer a Claude Agent run while it works (#2783).
  • Run a computer vendor with no exec through the dial host mode (#3013, #3019), and Mode 2 on a self-hosted Daytona
    target (#3021).
  • Read how 25 sandbox vendors fit the provider contract before writing a provider (#2994).

Fixed

  • A learned workflow's draft no longer answers a run (#3008).
  • SQLite: a wait for the write lock no longer fails StoreBusy when this process's own holder let go first (#3002).
  • Modal refused the exec pre-check's text check (about 66 KB of argv past its 64 KiB cap); every pre-check command now
    stays under 32 KiB (#3009).
  • Modal's kill() left the exec's process running; it now ends it and every process under it (#3001).
  • A run's learning that settled while it was read (GET /runs/{id}/learning, GET /sdk-runs/{id}) read merged or
    dropped with the proposal's time as its at once, then the learning run's end on every read after; it reads the end
    every time (#3014).
  • A run with skills, its folder on Archil and a computer that mounts it failed at its claim with 409 OperationAborted … another client holds a write delegation; the claim lays the skills (and a learning run's seed,
    which failed silently) before the computer mounts the folder (#2794).
  • Computer vendors' keys set where the server started stayed in its environment, readable in /proc/<pid>/environ
    and inherited by its children (#2777).
  • A webhook delivery whose name lookup was out when the server stopped could end the process with an unhandled
    rejection (#3024).
  • GET /v2/computers could list a paired computer online with seen_at: null (#3027).
  • A template whose feed trigger's urls defaults to [] failed validation though apply skipped it (#3023).
  • A url MCP need for a preset toolkit opened an empty add form under the toolkit's name; it opens the preset filled
    (#3003).

Also in this release

  • Computers: prepareImage's registry and vendor-side Dockerfile paths (#3000); sse beside ndjson in
    hosted/rest.ts, with Hyperbrowser's process API as PROVIDERS.md's worked example (#2996); one kill-by-pid for every
    exec with no kill of its own, Daytona on framedExec, an events stream that throws fails wait() (#3001); a deleted
    machine reads undefined from get and daemon: "child" says what PROVIDERS.md says (#2992); the label and lifetime
    trims (#2998); the vendor scorecard (#2994), re-scored after the trims: 0 clean, 5 fit if, 13 need a workaround, 10
    blocked (#3010).
  • Docs: the inbox quickstart applies its template with the plain Apply (#3004).
  • Tests and CI: the scripts tests that commit in a temporary repository run git with no background maintenance and
    remove it with retries, where a detached maintenance once dropped a merge queue run (#3007); the run-end test that
    cancelled a live run is gone, covered by the first-terminal-read case (#2995); CI checks every relative link and
    #anchor in the tracked Markdown files (scripts/check-doc-anchors.mjs) (#3010); the webhook-secret Timeline test
    looks for 8 characters of the secret in each form, 16 digits of its hex, where 8 digits matched a page cursor by chance
    about once in 50,000 runs, and a failure says where the piece showed (#3018).

Known limits

  • The learning loop is further along but not finished: a learned workflow that answers wrong with confidence is not caught by the money rule.
  • deploy/compose.yaml (a server and runners) has no Docker socket, so its agents get file tools and web_fetch but no shell; for a Docker sandbox, use the root compose.yaml, run the server on the host, or use a Daytona or E2B computer.
  • The other limits listed for beta.32 still hold.