fullsend-ai · waynesun09 · Aug 23, 2026 · Aug 23, 2026 · Aug 23, 2026 · Aug 23, 2026
@@ -640,6 +640,9 @@ jobs:
           TRIAGE_TARGET_REPO_DIR: target-repo
           TRIAGE_ANTHROPIC_VERTEX_PROJECT_ID: ${{ secrets.FULLSEND_GCP_PROJECT_ID }}
           TRIAGE_CLOUD_ML_REGION: ${{ inputs.gcp_region }}
+          # Per-run runtime/model/effort overrides from repository variables
+          # (FULLSEND_MODEL, TRIAGE_FULLSEND_MODEL, ...); see docs/runtimes.md.
+          FULLSEND_REPO_VARS: ${{ toJSON(vars) }}
         run: bash .github/scripts/setup-agent-env.sh
 
       - name: Run triage agent
@@ -755,6 +758,9 @@ jobs:
           CODE_ANTHROPIC_VERTEX_PROJECT_ID: ${{ secrets.FULLSEND_GCP_PROJECT_ID }}
           CODE_CLOUD_ML_REGION: ${{ inputs.gcp_region }}
           CODE_ISSUE_NUMBER: ${{ fromJSON(needs.route.outputs.event_payload).issue.number }}
+          # Per-run runtime/model/effort overrides from repository variables
+          # (FULLSEND_MODEL, CODE_FULLSEND_MODEL, ...); see docs/runtimes.md.
+          FULLSEND_REPO_VARS: ${{ toJSON(vars) }}
         run: bash .github/scripts/setup-agent-env.sh
 
       - name: Run code agent
@@ -877,6 +883,9 @@ jobs:
           REVIEW_TARGET_REPO_DIR: target-repo
           REVIEW_ANTHROPIC_VERTEX_PROJECT_ID: ${{ secrets.FULLSEND_GCP_PROJECT_ID }}
           REVIEW_CLOUD_ML_REGION: ${{ inputs.gcp_region }}
+          # Per-run runtime/model/effort overrides from repository variables
+          # (FULLSEND_MODEL, REVIEW_FULLSEND_MODEL, ...); see docs/runtimes.md.
+          FULLSEND_REPO_VARS: ${{ toJSON(vars) }}
         run: bash .github/scripts/setup-agent-env.sh
 
       - name: Run review agent
@@ -1143,6 +1152,9 @@ jobs:
           FIX_HUMAN_INSTRUCTION: ${{ steps.context.outputs.instruction }}
           FIX_FIX_ITERATION: ${{ steps.context.outputs.iteration }}
           FIX_REPO_FULL_NAME: ${{ github.repository }}
+          # Per-run runtime/model/effort overrides from repository variables
+          # (FULLSEND_MODEL, FIX_FULLSEND_MODEL, ...); see docs/runtimes.md.
+          FULLSEND_REPO_VARS: ${{ toJSON(vars) }}
         run: bash .github/scripts/setup-agent-env.sh
 
       - name: Run fix agent
@@ -1246,6 +1258,9 @@ jobs:
           RETRO_TARGET_REPO_DIR: target-repo
           RETRO_ANTHROPIC_VERTEX_PROJECT_ID: ${{ secrets.FULLSEND_GCP_PROJECT_ID }}
           RETRO_CLOUD_ML_REGION: ${{ inputs.gcp_region }}
+          # Per-run runtime/model/effort overrides from repository variables
+          # (FULLSEND_MODEL, RETRO_FULLSEND_MODEL, ...); see docs/runtimes.md.
+          FULLSEND_REPO_VARS: ${{ toJSON(vars) }}
         run: bash .github/scripts/setup-agent-env.sh
 
       - name: Run retro agent
@@ -1329,6 +1344,9 @@ jobs:
           PRIORITIZE_PROJECT_NUMBER: ${{ inputs.project_number }}
           PRIORITIZE_ANTHROPIC_VERTEX_PROJECT_ID: ${{ secrets.FULLSEND_GCP_PROJECT_ID }}
           PRIORITIZE_CLOUD_ML_REGION: ${{ inputs.gcp_region }}
+          # Per-run runtime/model/effort overrides from repository variables
+          # (FULLSEND_MODEL, PRIORITIZE_FULLSEND_MODEL, ...); see docs/runtimes.md.
+          FULLSEND_REPO_VARS: ${{ toJSON(vars) }}
         run: bash .github/scripts/setup-agent-env.sh
 
       - name: Create empty target-repo directory
@@ -1620,6 +1638,9 @@ jobs:
           MATRIX_ROLE: ${{ matrix.role }}
           GCP_PROJECT: ${{ secrets.FULLSEND_GCP_PROJECT_ID }}
           GCP_REGION: ${{ inputs.gcp_region }}
+          # Per-run runtime/model/effort overrides from repository variables
+          # (FULLSEND_MODEL, <ROLE>_FULLSEND_MODEL, ...); see docs/runtimes.md.
+          FULLSEND_REPO_VARS: ${{ toJSON(vars) }}
         run: |
           set -euo pipefail
           ROLE_UPPER=$(echo "${MATRIX_ROLE}" | tr '[:lower:]' '[:upper:]')

@@ -193,6 +193,7 @@ script-test:
 	$(call run-timed,bash .github/scripts/check-fix-eligibility-test.sh)
 	$(call run-timed,bash internal/scaffold/fullsend-repo/scripts/reconcile-repos-test.sh)
 	$(call run-timed,bash internal/scaffold/fullsend-repo/scripts/pre-fetch-prior-review-test.sh)
+	$(call run-timed,bash internal/scaffold/fullsend-repo/.github/scripts/setup-agent-env-test.sh)
 	$(call run-timed,bash hack/gitlab-runner-vm/executor/prepare_validation_test.sh)
 	$(call run-timed,python3 skills/topissues/scripts/topissues_test.py)
 	$(call run-timed,python3 skills/nextwork/scripts/nextwork_test.py)

@@ -185,6 +185,7 @@ export default defineConfig({
             { text: "fullsend inference", link: "/cli/inference" },
             { text: "fullsend mint", link: "/cli/mint" },
             { text: "fullsend repos", link: "/cli/repos" },
+            { text: "fullsend run", link: "/cli/run" },
           ],
         },
       ],
@@ -195,12 +196,17 @@ export default defineConfig({
           link: "/guides/getting-started/",
           items: [
             { text: "Getting Inference", link: "/guides/getting-started/getting-inference" },
+            { text: "Choose a Runtime", link: "/guides/getting-started/choosing-a-runtime" },
             { text: "Configuring GitHub", link: "/guides/getting-started/configuring-github" },
             { text: "Per-Org Mode", link: "/guides/getting-started/org-mode" },
             { text: "Repo Management", link: "/guides/getting-started/repo-management" },
             { text: "Operations", link: "/guides/getting-started/operations" },
           ],
         },
+        {
+          text: "Runtimes",
+          link: "/runtimes",
+        },
         {
           text: "Agents",
           collapsed: true,
@@ -244,7 +250,6 @@ export default defineConfig({
           items: [
             { text: "Vision", link: "/vision" },
             { text: "Architecture", link: "/architecture" },
-            { text: "Runtimes", link: "/runtimes" },
             { text: "Glossary", link: "/glossary" },
           ],
         },

@@ -24,7 +24,7 @@ Download the latest binary from [GitHub Releases](https://github.com/fullsend-ai
 
 | Command | Description |
 |---------|-------------|
-| `fullsend run` | Execute an agent locally in a sandbox. See [running agents locally](../guides/user/running-agents-locally.md). |
+| [`fullsend run`](run.md) | Execute an agent locally in a sandbox. See [running agents locally](../guides/user/running-agents-locally.md). |
 | `fullsend lock [agent-name]` | Pin remote dependencies to `lock.yaml` |
 | `fullsend scan` | Run security scanners on agent input/output |
 | `fullsend eval-measure` | Score wild-run traces into `eval-measurements.jsonl`. See [Eval measurements](../guides/infrastructure/eval-measurements.md). |

@@ -87,7 +87,7 @@ required in this case. This enables a greenfield setup without running
 
 Runs in three phases:
 
-1. **Manifest add** — repos specified as positional arguments that are not already in the manifest are added (`--forge` is required when the target platform cannot be inferred). Per-repo overrides (`--inference-region`, `--fullsend-ref`, `--mint-url`, `--allowed-remote-resources`) are written to the manifest entry.
+1. **Manifest add** — repos specified as positional arguments that are not already in the manifest are added (`--forge` is required when the target platform cannot be inferred). Per-repo overrides (`--inference-region`, `--fullsend-ref`, `--mint-url`, `--allowed-remote-resources`, `--runtime`) are written to the manifest entry.
 2. **Provision** — repos in the manifest that are not yet provisioned are installed (scaffold files, variables, secrets). Repos with a guard variable set but other components missing are repaired automatically.
 3. **Convergence** — repos that are already installed are checked for component drift (workflow, thin callers, variables, secrets — repaired automatically) and scaffold ref drift (upgraded automatically).
 
@@ -118,6 +118,7 @@ When repos are specified as positional arguments, only those repos are processed
 | `--fullsend-ref` | | Per-repo fullsend workflow ref override |
 | `--mint-url` | | Per-repo mint URL override |
 | `--allowed-remote-resources` | | Per-repo allowed remote resources override |
+| `--runtime` | | Agent runtime (`claude`, `pi`) recorded for repos this command adds; existing entries keep their `runtime` / `defaults.runtime` |
 | `--gitlab-bot-token` | | GitLab bot PAT for free-tier instances that don't support project access tokens (env: `FULLSEND_GITLAB_BOT_TOKEN`) |
 
 ### GitLab bot token
@@ -255,6 +256,7 @@ fullsend repos set-default github.mint_url ""   # removes the key
 | Key | Type | Description |
 |-----|------|-------------|
 | `defaults.allowed_remote_resources` | comma-separated URLs | HTTPS URLs agents may fetch at runtime |
+| `defaults.runtime` | `claude` or `pi` | Agent runtime written as each repo's `runtime:` at install; a per-entry `runtime` overrides it (`none` stops the chain) |
 | `github.url` | URL | GitHub instance URL (default: `https://github.com`) |
 | `github.mint_url` | URL | Token mint service URL (defaults to `https://mint.fullsend.sh` in public mode) |
 | `github.mint_mode` | `public` or `private` | Controls the default mint URL: `public` defaults to `https://mint.fullsend.sh`; `private` requires an explicit `mint_url` (default: `public`) |

@@ -0,0 +1,89 @@
+---
+sidebar_label: fullsend run
+---
+
+# fullsend run
+
+Execute an agent locally in a sandbox. `fullsend run` resolves the agent harness, provisions a sandbox container, and runs the agent to completion.
+
+## Usage
+
+```bash
+fullsend run <agent-name> [flags]
+```
+
+## Flags
+
+| Flag | Description |
+|------|-------------|
+| `--fullsend-dir` | Path to the `.fullsend` configuration directory |
+| `--runtime` | Override the agent runtime from `config.yaml` for this run (`claude`, `pi`, `dummy`); also `FULLSEND_RUNTIME` |
+| `--model` | Override the harness/agent model for this run (alias, model id, or `provider/id` on pi); also `FULLSEND_MODEL` |
+| `--effort` | Override the harness effort level for this run (`low`…`max`); also `FULLSEND_EFFORT` |
+| `--output-dir` | Base directory for run output (default: `/tmp/fullsend`) |
+| `--target-repo` | Path to the target repository |
+| `--fullsend-binary` | Path to a Linux fullsend binary to copy into the sandbox |
+| `--env-file` | Load environment variables from a dotenv file (repeatable) |
+| `--no-post-script` | Skip post-script execution |
+| `--keep-sandbox` | Skip sandbox deletion after the run |
+| `--debug [filter]` | Enable agent runtime debug logging with optional category filter (e.g. `"api,hooks"`) |
+| `--forge` | Forge platform to use (e.g. `"github"`, `"gitlab"`); auto-detected from CI env vars when omitted |
+| `--offline` | Reject network fetches; only use cached remote resources |
+| `--max-depth` | Maximum dependency depth for transitive resolution (0 disables) |
+
+## Plan block
+
+At startup, `fullsend run` prints a plan block summarizing the resolved configuration:
+
+```
+Agent:     code
+Role:      code
+Model:     sonnet
+Effort:    high
+Runtime:   claude (from /path/to/.fullsend/config.yaml)
+Image:     fullsend-sandbox:latest
+```
+
+The **Runtime** line shows which runtime was selected and the config source it was read from. When no `config.yaml` exists, the source reads `default (config not found)`.
+
+## Runtime selection
+
+The runtime for a run is resolved once, in this order: `--runtime` flag, `FULLSEND_RUNTIME`, the per-repo `runtime:` in `config.yaml` / `.fullsend/config.yaml`, then the built-in `claude`. The same order applies to the model (`--model`, `FULLSEND_MODEL`, harness `model:`, agent frontmatter; `FULLSEND_PI_MODEL` is a lower-precedence alias on pi) and to effort (`--effort`, `FULLSEND_EFFORT`, harness `effort:`). `FULLSEND_FALLBACK_MODELS=a,b` becomes Claude Code's `--fallback-model`; pi ignores it with a warning.
+
+The plan block prints `Runtime: <name> (from <source>)` and, when an override applied, `Model: <value> (from <source>)`; stderr carries `runtime: selected "<name>" from <source>` (and `model: requested "<value>" from <source>`) for scripts. An invalid override (unknown runtime, unknown effort level) fails before the sandbox is created.
+
+```bash
+# try a repo's triage on pi with Gemini Flash, without touching its config
+fullsend run triage --fullsend-dir . --target-repo ../repo \
+  --runtime pi --model google-vertex/gemini-2.5-flash --effort medium
+```
+
+## Output artifacts
+
+Each run produces artifacts in the output directory:
+
+| File | Description |
+|------|-------------|
+| `metrics.json` | Behavioral metrics: tokens, cost, model, runtime, iterations |
+| `transcripts/` | Agent conversation transcripts |
+| `claude-debug.log` or `pi-debug.log` | Debug log (when `--debug` is set) |
+
+### metrics.json fields
+
+| Field | Description |
+|-------|-------------|
+| `runtime` | Runtime that executed the run (e.g. `claude`, `pi`) |
+| `model` | Model the provider reported using |
+| `requested_runtime` | Runtime selected for the run (config file, or a `--runtime`/`FULLSEND_RUNTIME` override) |
+| `requested_model` | Model the harness/agent requested |
+| `override_source` | Where `requested_model` came from (`--model flag`, `FULLSEND_MODEL`, `FULLSEND_PI_MODEL`, `harness`, `default`) |
+| `runtime_source` | Where `requested_runtime` came from (`--runtime flag`, `FULLSEND_RUNTIME`, the config file path, or `default (config not found)`) |
+| `total_cost_usd` | Total inference cost |
+| `num_turns` | Number of conversation turns |
+| `iterations` | Number of retry iterations |
+
+## Related
+
+- [Running Agents Locally](../guides/user/running-agents-locally.md) for a step-by-step walkthrough
+- [Runtimes](../runtimes.md) for runtime selection and capabilities
+- [CLI internals](../guides/dev/cli-internals.md) for the full command tree
@@ -83,7 +83,7 @@ The `admin` command group's `install`/`uninstall`/`analyze`/`enable`/`disable` s
 
 | Category | Files |
 |----------|-------|
-| CLI reference | _(no dedicated page)_ |
+| CLI reference | `docs/cli/run.md` |
 | Guides | `docs/guides/user/running-agents-locally.md`, `docs/guides/user/building-custom-agents.md`, `docs/guides/dev/cli-internals.md` |
 | ADRs | `docs/ADRs/0036-agent-execution-sandbox.md` |
 | Contributing | `docs/contributing/sandbox-topology.md` |

@@ -0,0 +1,58 @@
+---
+sidebar_label: Choose a Runtime
+---
+
+# Choose an agent runtime
+
+Fullsend supports multiple agent runtimes. A runtime is the program that runs inside the sandbox and drives the model — it owns the tool-use loop, hook wiring, and transcript format. The runner (fullsend) owns everything outside: sandbox lifecycle, credentials, metrics, and the verdict.
+
+## Available runtimes
+
+| Runtime | Description | When to use |
+|---------|-------------|-------------|
+| `claude` | Claude Code (production default) | Most deployments — mature, full sub-agent support |
+| `pi` | [Pi](https://github.com/earendil-works/pi) — Claude on Vertex by default; any provider pi supports by model name (e.g. Gemini on Vertex with the same credentials) | Opt-in alternative; no sub-agent tool yet, so `review`/`retro` run single-context; see [Runtimes](../../runtimes.md) for known constraints |
+
+## How to select a runtime
+
+### At setup time
+
+`fullsend github setup <owner/repo>` asks which runtime to use when run from a terminal (press Enter for `claude`), and the setup PR it opens describes the choice. To set it explicitly, pass `--runtime`:
+
+```bash
+fullsend github setup <owner/repo> \
+  --inference-wif-provider "<wif-url>" \
+  --runtime pi
+```
+
+### In config
+
+Set `runtime:` in the repo's `.fullsend/config.yaml`:
+
+```yaml
+runtime: pi
+```
+
+For fleets managed through `repos.yaml`, set `defaults.runtime` (or a per-entry `runtime`) and run `fullsend repos install`, or pass `fullsend repos install --runtime pi` for repos the command adds:
+
+```bash
+fullsend repos set-default defaults.runtime pi
+```
+
+To try a runtime or model on one run without changing the repo, use the per-run overrides — `fullsend run --runtime pi --model google-vertex/gemini-2.5-flash`, or the `FULLSEND_RUNTIME` / `FULLSEND_MODEL` / `FULLSEND_EFFORT` environment variables (flag beats environment beats config).
+
+No workflow file change is needed — the runner reads the config at the start of every run.
+
+## Where to see what ran
+
+After a run completes, the selected runtime and model appear in several places:
+
+- **Run plan block** — `Runtime: <name> (from <source>)` printed at the start of every `fullsend run`
+- **Status comment** — the terminal status comment on the issue/PR includes a footer with runtime, model, effort, and cost
+- **metrics.json** — `runtime`, `requested_runtime`, `runtime_source`, `requested_model`, and `override_source` fields record what was selected and why
+- **stderr** — `runtime: selected "<name>" from <source>` for script consumers
+
+## Next steps
+
+- [Configuring GitHub](configuring-github.md) to set up your repo
+- [Runtimes](../../runtimes.md) for the full runtime reference, including model override precedence and the capability table
@@ -112,6 +112,8 @@ fetched content before committing.
 > because the preset provides its own configuration. `--config` is only
 > valid for per-repo mode.
 
+The `--runtime` flag selects which agent runtime to use (e.g. `claude` or `pi`); see [Choose a Runtime](choosing-a-runtime.md) for what the runtimes are and how to change the selection after setup.
+
 ## Testing Fullsend
 
 After installing open a new issue or comment `/fs-triage` in an open issue. Then visit the

@@ -572,7 +572,9 @@ Stop reasons: toolUse=2, stop=1
 
 | Variable | Description |
 |----------|-------------|
-| `FULLSEND_PI_MODEL` | Override the model id (runner env) |
+| `FULLSEND_MODEL` (or `fullsend run --model`) | Override the model for the run on any runtime; `FULLSEND_PI_MODEL` is kept as a pi-only alias |
+| `FULLSEND_RUNTIME` (or `--runtime`) | Override the runtime selected by `config.yaml` |
+| `FULLSEND_EFFORT` (or `--effort`) | Override the harness effort level |
 | `FULLSEND_PI_PROVIDER` | Override the inference provider (runner env) |
 | `FULLSEND_PI_BASH_ALLOWLIST` | Set to `enforce` to make the Bash first-token allowlist block instead of warn |