Runners: agents inside your own network
A runner runs your organization's AI agents inside your own network: your cloud account or your own data center. You still build and manage agents in the SourceLace app, but their runs happen on the runner. Your data, the AI calls, everyone's source sign-ins and the results never leave your network; only each run's status, steps and counts go back to SourceLace.
A runner runs only your AI agents. Everything else, including the SourceLace app, stays on SourceLace's cloud; SourceLace itself is not offered for installing in your own network.
Each agent can run on SourceLace's cloud or on a runner, so you can use both: see Cloud and runners together (hybrid).
Who can have a runner
Runners are offered on the Enterprise plan only, on request. They come with an annual runner fee on top of your subscription, a one-time onboarding fee (installation help and a security review), Premium support (faster response times, named contacts and a scheduled upgrade plan) and a minimum contract term of one year. Prices are not published: ask your SourceLace contact.
SourceLace applies these terms itself: runners can only be switched on for an organization on the Enterprise plan, and a runner can only be added while your organization has Premium support and runners on its contract. Your SourceLace contact sets what your contract includes; your admins add runners within it.
Your SourceLace contact switches runners on for your organization. When they are on, admins see a Runners page in the app.
Cloud and runners together (hybrid)
You choose where each agent runs: on SourceLace's cloud or on one of your runners. One organization can do both at once, often called hybrid. An agent that needs a system SourceLace's cloud cannot reach, such as a database with no internet address, runs on a runner (What your network needs).
An example. Corvanta keeps its sales data in Salesforce and its payroll in a database in its own data center that may not be reached from outside.
- Its pipeline review agent reads Salesforce. It runs on SourceLace's cloud, so nothing needs to be installed for it.
- Its payroll check agent reads the payroll database. It runs on Corvanta's runner in the data center, so the payroll rows, the agent's summary and its results never leave Corvanta's network.
Both agents are built, approved and watched in the same SourceLace app, by the same people.
Choosing for each agent. In the agent editor, open the Results tab and pick under Where it runs: SourceLace (the cloud, the default) or one of your organization's runners. The choice belongs to the agent, so each agent can run in a different place.
Moving an agent. Edit the agent, choose the other place under Where it runs and save. Its next run happens in the new place. Before that run:
- the agent's owner connects the agent's sources in the new place, because sign-ins are kept where the agent runs (Your sources on the runner);
- check that the agent uses only what the new place offers (see "Things to know" below); saving tells you if it does not.
Earlier runs stay where they ran: the results of runs on a runner stay on that runner.
What people see in the app. One Agents list holds every agent, wherever it runs. An agent on a runner says so on its page, and each of its runs shows Ran in your network with an Open result link to the runner's page (Running an agent on a runner). The run history, schedules, versions and audit trail are in the app for every agent. Changes proposed by agents on SourceLace's cloud wait on the app's Approvals page; changes proposed by agents on a runner wait on that runner's page, so their data stays in your network (Approvals on the runner).
Requiring runners for every agent. An admin can turn on Every agent runs on one of our runners under Agents → Settings. From then on, an agent can be saved or switched on only when it names one of your runners. The switch does not move agents that already run on SourceLace's cloud: they keep running there until someone edits them or switches them on again, which then needs a runner. To move them at once, choose a runner for each one under Where it runs.
Things to know.
- Runners are on the Enterprise plan only, on request (Who can have a runner). Without a runner, every agent runs on SourceLace's cloud.
- Sign-ins are kept per place. A person's sign-ins on SourceLace's cloud are not on a runner, and each runner keeps its own. Each person connects their sources once on each runner their agents use.
- Results stay where the agent ran. A run on a runner keeps its summary, files, drafts and changes waiting for approval on the runner; SourceLace gets only its status, steps and counts (What stays where).
- Chat and AI apps always use SourceLace's cloud. A runner runs only agents. Questions asked in the SourceLace app, and AI apps connected to SourceLace, use SourceLace's cloud and the sign-ins people made there.
- Some agent features are on SourceLace's cloud only. On a runner, an agent starts on a schedule, with Run now, or from its webhook or email address. Calendar and event triggers, starting after other agents, asking or being asked by other agents (including outside agents), sending the files it makes, and routing decisions to groups are available only for agents on SourceLace's cloud. An email's attachments are kept only for agents on the cloud (Before you choose a runner).
Before you choose a runner: what to weigh
A runner keeps your agents' data inside your network. It also changes a few things compared with agents on SourceLace's cloud, so it is worth weighing them before you decide:
- New versions reach a runner when your IT department installs them. On SourceLace's cloud, fixes (including security fixes) and new features arrive as soon as SourceLace releases them. A runner gets them only when your IT department installs the new version, so they reach runs on the runner later. An out-of-date runner keeps working, but may refuse runs that use rules newer than it can read until it is upgraded. Your SourceLace contact tells you about new versions and which ones are supported, and Premium support includes a scheduled upgrade plan (Upgrading).
- Some features are not available on runners yet. Agents on a runner start on a schedule, with Run now, or from their webhook or email address; calendar and event triggers are not available, and an email's attachments are not kept for a runner run, only their names and sizes. A change that an approval rule says someone in a group must approve cannot be approved on the runner's page. A quiet runner is shown to admins in the app, not by email.
- You run the infrastructure. Your organization provides and pays for the machine or cluster, the database, the disk and the network, and is responsible for patching the host, backing up the database, keeping the master key safe and keeping the runner's page inside your network (Who is responsible for what). What is needed is listed in Runner requirements and planning.
- Capacity and uptime of runner runs depend on your runner. Runs on a runner run as fast, as many at a time and as reliably as the machines and network you give it. Run it as two or more instances on different machines and one can fail without anyone doing anything: another takes its runs over (Failover). When every instance is stopped or cannot reach SourceLace, runs due on it wait; nothing is lost, and admins see a notice.
- Cost. A runner comes with an annual runner fee on top of your subscription, a one-time onboarding fee (installation help and a security review), Premium support and a minimum contract term of one year (Who can have a runner).
SourceLace helps with each of these: onboarding helps your IT department install the runner, Premium support gives you named contacts and the upgrade plan, and the runner's own connectivity check and support bundle make problems quick to find (Runner troubleshooting).
Production and Test runners
Each runner is a Production or a Test runner. Your contract says how many of each your organization has; a Test runner costs less, for trying new agents and versions before they run your real work. The Runners page shows, for each label, how many runners are in use and how many your contract includes.
An admin chooses the label when adding a runner, and can add runners of a label until the contract's number is reached. To add more, ask your SourceLace contact. Revoking a runner frees its place. A runner's label does not change: to move one, add a new runner with the other label and revoke the old one.
What stays where
| On the runner, in your network | On SourceLace |
|---|---|
| Every run: what the agent read, its summary, the files it made, drafts and changes waiting for approval | The agent itself: its name, instructions, sources, actions and schedule |
| Each person's sign-ins to your sources | Who owns each agent, and where it runs |
| Your organization's own AI key, and every call to the AI model | Each run's status, times, one line per step (tool, source, object, row count, outcome) and counts |
| Secret options of your sources, such as client secrets, and the OAuth apps registered for the runner | The audit trail of what each runner asked for and reported |
A run's report to SourceLace has no place for a summary, rows, record values, drafts or files, and SourceLace refuses any report that tries to send one.
Your rules on the runner
Your admins set the rules once, in SourceLace, and they apply the same way to agents on a runner as to agents on SourceLace's cloud:
- Data protection rules: the fields your admins hide or mask, and the kinds of value SourceLace masks wherever they appear (such as email addresses), are hidden or masked before the AI model sees them. People in a group a rule names see the real values on the runner's page, as they would in SourceLace.
- Access by group: an agent reads and changes only the sources and objects the groups of the person it runs for allow, and is refused the rest with the same message as in SourceLace. This never hides a source from the runner's Your sources page: anyone signed in to the runner can connect the sources its agents use.
- Read rules: an agent reads only the rows the groups of the person it runs for may read.
- Approval rules and fixed values for changes.
- Systems switched off for your organization: if SourceLace switches a system off for your organization, or a kind of system needs a feature your organization does not have, agents on a runner cannot use it either. A run of an agent that uses it does not start: it ends as stopped, with the system named in its history. A run already going stops at its next check-in. The runner also forgets that system's settings, including any secret settings an admin entered on the runner's Settings page; when it is switched on again, an admin enters those again there. SourceLace checks this itself, so it applies whatever happens on the runner.
With each run, SourceLace sends the runner these rules for that agent's sources only. They are settings, never your data, and the runner keeps them in memory, not in its database. To change them, change them in SourceLace: the agent's next run uses the new rules. The runner never runs an agent without them: if a run arrives without your rules, or the runner is too old to read them, the run is marked as failed before it reads anything, with the reason in its history. Update the runner to its latest version, then run the agent again.
If the runner restarts while a change waits for approval, approving it right away says the runner does not have the person's access settings yet, and nothing is written. The change keeps waiting: approve it again after that person's next agent run on the runner.
Setting one up
Your IT department installs the runner inside your network, with its own database, and connects it to SourceLace. The complete guide for them is in four chapters: Runner requirements and planning (where it can run, sizing, database, every network connection, skills and checklists), Installing a runner, Runner troubleshooting and Looking after a runner. In short:
- An admin opens Runners in the app, names the runner (such as "Frankfurt VPC"), chooses Production or Test, presses Add a runner and copies the token. It is shown once.
- IT starts the runner from the private, versioned image SourceLace made for your organization, which an admin gets from Runners, Get the runner (Get the runner), with its own database, a master key only you hold, the token and the runner page's own https address (such as
https://runner.corvanta.internal). - The runner connects out to SourceLace over HTTPS and asks for work about every 15 seconds. You open no inbound port; its page is only for people inside your network.
- An admin signs in to the runner's page and, under Settings, enters your organization's own AI key (Anthropic or OpenAI) and any secret source options.
The runner's license
A runner needs a current license from SourceLace. It gets one by itself, during its usual contact with SourceLace, and it is renewed automatically while your contract includes runners of its kind (Production or Test). Nobody installs a license by hand. The runner's Settings page shows its license and when it expires, and the Runners page in the app shows the same for each runner.
A license lasts up to 30 days and never past your organization's runner license end date: the end date of your contract, unless your contract renews by itself, or another date agreed with your SourceLace contact. From 30 days before that date, the Runners page and a notice at the top of the app show "Your runner licenses end on ...". Extending your contract renews the licenses at the runners' next contact (License end dates).
Without a current license, for example when the contract no longer includes the runner, the runner takes no new work. Runs already in progress finish, changes waiting for approval can still be decided, and nothing is deleted. Once the license is renewed, the runner picks up work again on its next contact. A runner that cannot reach SourceLace for a long time (about a month) also stops taking new work until it is back in touch.
Signing in to the runner
Open the runner's page inside your network and press sign in. Your browser goes to SourceLace, where you sign in as usual, and comes straight back. SourceLace tells the runner only your email address, your organization and your role. Only members of the runner's organization can sign in, and never to a revoked runner. A sign-in on the runner lasts 12 hours.
Your sources on the runner
Agents act as their owner, with the owner's own sign-ins, exactly as on SourceLace. Those sign-ins are kept on the runner, so each person connects their sources once on the runner's page, under Your sources. The runner shows the sources your organization's agents on that runner use.
Where a source signs in through an OAuth app your organization registers itself (such as ServiceNow, Workday, Oracle Fusion, Snowflake or Salesforce), register it with the runner's own callback address, such as https://runner.corvanta.internal/connect/callback, and an admin enters its secret on the runner's Settings page. SourceLace never receives those secrets.
Running an agent on a runner
In the agent editor, choose the runner under Where it runs, on the Results tab. An admin can also require that every agent runs on one of your runners. Such an agent starts on a schedule, with Run now, or from its webhook or email address (an email's attachments are not kept for a runner run, only their names and sizes). Calendar and event triggers are not available on runners yet, because checking them needs the owner's sign-ins, which are on the runner.
When a run is done, its run page on SourceLace shows Ran in your network and an Open result link to the runner's page, where the summary, the steps and the files are.
Approvals on the runner
Changes an agent proposes wait under Approvals on the runner's page. The agent's owner or an admin approves or rejects each one there. An approved change is made at once, as the agent's owner, with the owner's sign-in held on the runner, and with your organization's rules as they are at that moment (the runner asks SourceLace for them first). Once a run has nothing left to approve, SourceLace hears that it is done.
How long results are kept
Runs, their files and decided approvals are deleted from the runner after your organization's retention period, which an admin sets on the runner's Settings page (30 days unless you change it).
Is the runner healthy?
The Runners page shows each runner's status, when it last asked for work, its version and its health: how long it has been running, how many runs it holds, the type of its last error, free disk in the folder where it keeps large results, and free memory. The runner sends only these counts, never messages or data.
- Healthy: it asks for work as usual.
- Quiet: it has not asked for work in the last 90 seconds. It may be stopped, or unable to reach SourceLace.
- Out of date: it works, but a newer runner version exists. Ask IT to update its image.
- Revoked: its token no longer works.
A runner that runs as two or more instances (Failover) lists them under its health: each instance's name and version, and how many are asking for work, such as "2 of 3 instances asking for work". The runner shows Healthy while any of them asks for work. An instance not seen for a day leaves the list.
Under its version, each runner shows its support: each version is supported until 6 months after the next version is released. Update available (a newer version exists), Upgrade due (less than a month of support left) or Unsupported (it takes no new work until it is upgraded). See How long each version is supported.
Under its status, each runner shows until when it is licensed, or that its license expired and it takes no new work.
When a runner with agents assigned to it stays quiet for 10 minutes, admins see a notice at the top of the app until it is back, and the audit trail records it once. An admin can shorten the 10 minutes (to as little as 2) on the Limits page. Runs due on a quiet runner wait for it, and their run history says so.
If the runner restarts or loses its connection
Nothing is lost and no run is done twice. Until SourceLace has a run's report, the runner keeps it in its own database and sends it again, also after a restart. If a run is stopped halfway, by a restart or a long network outage, it starts again from scratch only when it had not yet made any draft, change, post or message. If it had, it is marked as failed with "Interrupted after it had made drafts or changes, so it was not run again": check what it made on the runner's page, then run it again yourself if needed. Keep your runner on the latest version: older runners cannot tell SourceLace in time, so their interrupted runs that could have made changes are not run again either. With two or more instances, the same rules hold when one instance stops: another takes the run over after about three minutes, and never makes a change twice (Failover).
When something is wrong
On the runner's own Settings page, an admin can:
- Check connectivity: whether the runner reaches SourceLace (through your proxy, if
HTTPS_PROXYis set), its own database, and each of your sources. Sources are only checked for whether their address answers: no sign-in is used and nothing is read. Every problem comes with what to do, such as asking your network team to open a firewall rule. - Download support bundle: a zip to send SourceLace support, with the runner's version, its settings with every secret left out, the types of recent errors and how many of each, and the connectivity check. It holds no tokens, keys, passwords, AI keys, data or run results, and nothing about people beyond counts.
Every message the runner can show, with what it means and how to fix it, is in Runner troubleshooting.
When runners get no new work
If your organization no longer meets the runner terms, its runners get no new work until it does again: when it is no longer on the Enterprise plan, no longer has Premium support, or has more runners of a label in use than its contract includes (for example after the contract was changed). Only the runners concerned stop getting work: with too many Production runners, Test runners keep working.
The same happens to a runner whose version is no longer supported (until it is upgraded), and to every runner after your organization's runner license end date (until your contract is extended).
- Runs already in progress finish as usual, and their results stay on the runner.
- No runner is removed, and nothing is deleted.
- Admins see a notice at the top of the app saying why, and the Runners page shows it next to each runner concerned. Runs due on those runners wait in the queue.
- When the terms are met again (for example after an admin revokes a runner, or your contract is updated), work starts again at the runner's next request.
At the end of your contract, your runner licenses end too, unless your contract renews by itself: after that day your runners take no new work until the contract is extended (runs in progress finish, nothing is deleted). Admins see the date from 30 days before. Ask your SourceLace contact about renewals.
Stopping a runner
Pausing or stopping an agent, and your organization's stop switch, reach the runner at its next request for work, within about 15 seconds: a run in progress stops at its next step. Revoking the runner on the Runners page stops its token at once: it gets no more work, nobody can sign in to it, and runs queued for it end as failed. A run already going on it is stopped too, at the runner's next contact with SourceLace: usually within about 15 seconds (its next request for work), and at most about a minute later (its next check-in for the run, which it makes every 60 seconds). The run is marked as failed with "SourceLace refused this runner's token ..." and writes nothing more to your systems; anything it made before is listed on the runner's page. From then on the runner applies no change waiting for approval (you can still reject them), its Settings page shows that SourceLace refused its token, and its log says how many runs it stopped. A network problem does not do this: runs keep going while the runner tries again. To end everything at once, your IT department also stops the runner (Revoking and removing a runner).
See also AI agents, Runner security and Security and privacy.