Runsight is live · free · last verified 25 Oct 2026 · viability monitor
Runsight
Open-source, Apache 2.0 YAML workflow engine for Git-native AI agents with per-block cost tracking and built-in evals.
Runsight is a practical choice for teams that already manage AI agents like software. Git-native YAML workflows, block-level cost tracking with max_cost and max_tokens caps, and a built-in assertion/eval harness give you real control and reproducibility, and `uvx runsight` gets you to localhost:8000 with no signup. The visual canvas + Monaco editor pairing means a teammate who prefers dragging blocks and one who prefers diffing YAML work on the same file. The trade-off is explicit: you run your own infrastructure and hold your own API keys, and outreach beyond HTTP, Python, and file-based custom tools is your job to wire. If you need a fully managed studio, look at managed agent platforms
Verdict generatedVerified 5d ago · liveness 60/100 · cite: rightaichoice.com/tools/runsight
- Developer teams building multi-step AI agent pipelines
- Teams that want agent behavior reviewable in pull requests
- Organizations needing per-run cost visibility and hard spend caps
- Engineers building eval and regression tests for agent outputs
- Non-technical users who need a no-code-only interface
- Teams that want a fully managed service with zero infrastructure work
- Buyers looking for a marketplace of ready-made agent templates
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Runsight if you want a hosted console that manages your infrastructure and API keys for you, or if you need pre-built connectors and a template gallery rather than agent logic you author as YAML.
Runsight is Apache 2.0 open source and runs on your own machine with your own model API keys, so the license itself costs nothing — your real spend is the model usage Runsight measures per block, plus the compute and engineer time you put into running it. Budget accordingly against managed agent platforms that bundle hosting into a subscription: you are trading subscription fees for infrastructure you operate, with max_cost and max_tokens caps as the guardrail against model spend surprises.
In short
Runsight — Open-source, Apache 2.0 YAML workflow engine for Git-native AI agents with per-block cost tracking and built-in evals. Best for Developer teams building multi-step AI agent pipelines, Teams that want agent behavior reviewable in pull requests, Organizations needing per-run cost visibility and hard spend caps. Free to use.
What is Runsight?
Runsight — Open-source, Apache 2.0 YAML workflow engine for Git-native AI agents with per-block cost tracking and built-in evals. Best for Developer teams building multi-step AI agent pipelines, Teams that want agent behavior reviewable in pull requests, Organizations needing per-run cost visibility and hard spend caps. Free to use.
Runsight is an open-source, Apache 2.0-licensed workflow engine that lets you design, version, and run AI agent pipelines as YAML files stored in Git. Instead of scattering agent logic across Python scripts, you define workflows in a readable YAML format that lives in your repository, so every change is diffable, reviewable, and revertible like any other code. Workflows are built from five block types — linear, gate, code, loop, and workflow — and reference reusable "souls": agent identities that carry a role, prompt, model, and tool list, stored as library files or inline. A key differentiator is the duality between a visual canvas and a Monaco-based YAML editor: both stay bidirectionally synced, so you can drag nodes and the YAML updates, or edit YAML and the canvas reacts. Every workflow is a file on disk and every save is effectively a commit, enabling Git-native version control with simulation branches and fork recovery. Runsight tracks per-run cost per block, including latency, and lets you set hard budget caps (max_cost and max_tokens) that kill execution before overspend. The built-in evaluation framework uses assertions on block outputs, with transform hooks for structured extraction and an offline eval test harness for regression testing across runs. You can pause, inspect, and resume or kill agents mid-run, and the engine supports loops with break conditions plus sub-workflows as nested child steps for hierarchical orchestration. Runsight is self-hosted: it runs on your machine with your own API keys and models, and nothing leaves your network unless you choose to send it. It is a fit for developer teams who already treat agent behavior as code and want to collaborate through branch, review, and merge cycles. Compared to fully-managed agent platforms, Runsight leaves infrastructure and API keys in your hands, but for Git-centric engineering teams it is a transparent, lock-in-free option where your workflows stay readable YAML even if the project disappears.
Key features
- YAML-first workflow definitions (version, entry, blocks)
- Five block types: linear, gate, code, loop, workflow
- Git-native version control — every save is a commit
- Visual canvas editor with drag-to-arrange blocks
- Monaco YAML editor bidirectionally synced with the canvas
- Per-run cost tracking per block, including block latency
- Hard budget caps (max_cost, max_tokens) that halt execution before overspend
- Built-in eval framework with block-level assertions
- Transform hooks for structured extraction
- Offline eval test harness for regression testing across runs
- Pause, inspect, and resume or kill agents mid-run
- Loops with break conditions
- Sub-workflows and nested hierarchical orchestration
- Souls — reusable agent identities (role, prompt, model, tools)
- Custom tools with canonical IDs via HTTP, Python, or file-based executors
Behind the verdict
Runsight's central bet is that an agent pipeline is a config artifact, not an application. That bet shows up everywhere in the product. A workflow is a YAML file at custom/workflows/<name>.yaml with a version field, a named entry block, and a map of blocks. Because it is a file, `git diff` works on it — the homepage's own example shows a diff that swaps soul_ref: analyst for soul_ref: senior-analyst and adds a max_cost: 0.50 budget. `git commit -m "upgrade analyst soul, add cost cap"` then `runsight run research-pipeline` is the whole loop. The block model is small and legible: linear, gate, code, loop, and workflow. Gates are where the eval story connects — a gate block with an eval_key routes pass to one block and fail back to another, so the homepage's Research Pipeline sends a failed review back to research rather than publishing. Souls keep agent definitions out of the graph: role, prompt, model, and tools live in a reusable identity you reference by soul_ref, which is exactly what makes the analyst-to-senior-analyst swap a one-line diff instead of a refactor across files. Tools are defined as YAML with canonical IDs and HTTP, Python, or file-based executors. The cost surface is the part most teams will feel immediately. Each block reports its own spend and latency (the homepage shows research $0.032 / 1.2s, review $0.018 / 0.8s, total $0.050 against a $2.00 cap), and budget caps halt execution before overspend rather than alerting you afterward. Pause-and-kill-mid-run is the same instinct applied to behavior: stop a drifting agent before it burns the rest of the cap, inspect state, then resume or kill. Where it asks something of you: the docs are organized as Quickstart, Key Concepts, YAML Schema, Block Types, Souls, Custom Tools, Git Integration, and Assertions and Eval — a real manual, but you will read it. Self-hosting means uptime, scaling, and secret management are yours. And the connector story is custom tools you write, so anything you need beyond HTTP, Python, and file executors is integration work in your repo. Compared with managed agent studios, you trade a hosted console and pre-built connectors for a repo you fully own and a format that outlives the vendor — the homepage is blunt about this: "If Runsight disappears, you still have readable config." For a platform team standardizing how three squads build agents, that is the argument. For a solo operator who wants to click together a chatbot today, it is friction with no payoff.
Models under the hood
as of 2026-10-24
Limitations
- Runsight is self-hosted: you manage your own infrastructure, uptime, scaling, secrets, and API keys, and nothing is handled for you.
- Tool connections are custom tools you define in YAML with HTTP, Python, or file-based executors, so external systems are integration work rather than pre-built connectors.
- The engine orchestrates workflows — it does not host or train models, and it ships no marketplace of ready-made agents.
- Expect a real reading curve through the docs (YAML Schema, Block Types, Souls, Custom Tools, Git Integration, Assertions and Eval) before your first non-trivial pipeline, and the community is still early, so GitHub issues and docs are your main support channel.
as of 2026-10-02
Researching Runsight? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is Runsight still active?
Runsight is live · free · last verified 25 Oct 2026 · viability monitor.
Viability score
How well maintained and how widely used is Runsight? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →What do users say about Runsight?
What people actually say about Runsight — is it worth it?
We scanned public community sources for Runsight on Oct 18, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Only 0 of the posts we fetched could be positively tied to Runsight. Rather than publish a sentiment score built on the wrong subject, we publish nothing here and re-run the scan.
How we verified this
Every listed fact about Runsight— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 25 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified Runsight 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed SSL certificate
- — changed description, our verdict, features, use cases, limitations, switching in, switching out, who should skip it, pricing tiers, pricing fit, setup time, summary, who it suits, hidden costs
- — changed domain expiry
- — changed hidden costs, who should skip it, pricing fit, switching in, features, switching out, setup time
- — changed SSL certificate
- — changed pricing page
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does Runsight cost in 2026?
Runsight: Free to use. Pricing last verified 25 Oct 2026.
Plans compared
Every published Runsight tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Open Source | Free | — | Developer teams that already use Git and are prepared to self-host Runsight with their own model API keys.Adds: Starting tier — the full Apache 2.0 source with no paid tier above it: YAML workflows, Git-native versioning, canvas + Monaco editor, per-run cost tracking, budget caps, and the eval framework are all included. |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where Runsight's pricing actually pencils out — and where peers do it cheaper.
Runsight is Apache 2.0 open source and runs on your own machine with your own model API keys, so the license itself costs nothing — your real spend is the model usage Runsight measures per block, plus the compute and engineer time you put into running it. Budget accordingly against managed agent platforms that bundle hosting into a subscription: you are trading subscription fees for infrastructure you operate, with max_cost and max_tokens caps as the guardrail against model spend surprises.
Has Runsight's pricing changed?
No plan price for Runsight has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick Runsight?
Pick Runsight if you are Developer teams building multi-step AI agent pipelines, Teams that want agent behavior reviewable in pull requests, Organizations needing per-run cost visibility and hard spend caps, Engineers building eval and regression tests for agent outputs.
Look elsewhere if you are Non-technical users who need a no-code-only interface, Teams that want a fully managed service with zero infrastructure work, Buyers looking for a marketplace of ready-made agent templates, Organizations that need pre-built connectors rather than YAML-defined custom tools.
Skip it if Skip Runsight if you want a hosted console that manages your infrastructure and API keys for you, or if you need pre-built connectors and a template gallery rather than agent logic you author as YAML.
GeneratedReal-world workflow fit
Concrete scenarios for the personas Runsight actually fits — and what changes day-one when you adopt it.
Installs with `uvx runsight`, lands on localhost:8000, and writes a first research-pipeline.yaml using linear blocks with soul_refs for analyst and editor, then commits it to the team repo.
Outcome: A shared, diffable workflow file other squads can branch from instead of three divergent Python scripts.
Adds a budget block with max_cost: 2.00 and max_tokens: 50000 to the workflow, then watches per-block cost and latency stream in during a run.
Outcome: Visibility into which block drives cost (for example research $0.032 / 1.2s versus review $0.018 / 0.8s) and automatic halt before overspend.
Puts a gate block with an eval_key on the review step, routing pass to publish and fail back to research, then runs the offline eval harness against the change.
Outcome: Failed drafts loop back for another pass instead of publishing, and the change is regression-tested before merge.
Use cases
- Version AI agent pipelines as YAML files inside the same Git repo as your application code.
- Track per-block cost and latency across a multi-step agent pipeline to find the expensive step.
- Set max_cost and max_tokens caps so a runaway loop stops before it overspends.
- Gate a publish step behind assertions and route failed reviews back for another research pass.
- Pause a running agent, inspect its state, then resume or kill it.
- Run an offline eval harness to catch regressions before merging a soul or prompt change.
- Share one soul definition (role, prompt, model, tools) across several workflows instead of duplicating prompts.
- Wire internal systems in as custom HTTP or Python tools instead of moving data to a SaaS console.
Setup time & first value
How long it actually takes to get something useful out of Runsight — broken out by persona, not the marketing-page minute.
Platform engineer: roughly 5 minutes to a running local instance via `uvx runsight`, then 30–60 minutes to a first meaningful multi-block workflow with souls and a budget cap. Quality-focused developer: about an hour to add a gate with an eval_key and run the offline eval harness. Expect longer for anything involving custom HTTP or Python tools, since you author those yourself.
Switching to or from Runsight
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From hand-rolled Python agent scripts: move each script into a workflow YAML — express the sequence as linear blocks and move prompt/model/tool choices into souls referenced by soul_ref.
- →From a notebook prototype: lift the chains into blocks, then commit the YAML so further prompt edits show up as diffs.
- →From another orchestration tool: re-express the graph with Runsight's five block types (linear, gate, code, loop, workflow) and re-add spend limits as max_cost and max_tokens.
- ↗To a managed agent platform: port each workflow YAML into the platform's hosted graph editor, since souls and block definitions do not import directly.
- ↗To your own framework code: read the workflow files and translate blocks to your runtime, then re-implement budget caps and assertions.
- ↗To another Git-native orchestrator: the YAML is plain enough to hand-translate, but gate eval_keys, transform hooks, and soul references need to be rebuilt in the target format.
Tutorials & Learning
YouTube returned 6 videos for “Runsight”, and we withheld 6: 6 could not be judged, because “Runsight” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Runsight.
Official links
What are the best Runsight alternatives?
The closest alternatives to Runsight we list are Inngest, Mastra and omnigent. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Inngest | Freemium | Pricing not published | — |
| Mastra | Freemium | Pricing not published | — |
| omnigent | Free | Free | — |
Alternatives to Runsight
View all
Tools that pair well with Runsight
Common stack mates teams adopt alongside Runsight, with the specific reason each pairing earns its keep.
Inngest
Open source durable execution SDK for background jobs, workflows, and AI agents — wrap code in step.run() and get retries, flow control,
Mastra
Open-source TypeScript agent framework with durable workflows, memory, and observability built into one stack.
omnigent
Open-source meta-harness that runs Claude Code, Codex, Pi and your own agents under one policy, sandbox and cost layer
Runsight vs its closest rival
Runsight's closest head-to-head is with Spider Cloud.
Our Runsight vs Spider Cloud verdict written Jul 2, 2026 is more than 90 days old, so it is being re-reviewed against current data and is not shown.
Frequently asked questions
Best-of guides
Used Runsight? Help shape our editorial sentiment research.